Introduction
Degenerative lumbar spine diseases are a leading cause of disability worldwide, often necessitating pedicle screw fixation and fusion to achieve clinically significant improvement [1]. Lumbar interbody fusion (LIF) involves placing a cage or graft between vertebral bodies to promote fusion, often supplemented with pedicle screw fixation. This study focuses on patients undergoing LIF for degenerative lumbar diseases, including posterior (PLIF), transforaminal (TLIF), lateral (LLIF), and anterior (ALIF) approaches. Screw loosening (SL) is a common complication following LIF, potentially resulting in nonunion, chronic pain, progressive kyphotic deformity, or even necessitating revision surgery [2]. The incidence of SL varies widely, ranging from 0.8% to 35%, with rates exceeding 60% in patients with osteoporosis [3–7].
SL is influenced by various factors, including advanced age [6], body mass index (BMI) [8], bone density [9], Cobb angle [10], long-segment fusion [11], inadequate restoration of lumbar lordosis (LL) [12], and insufficient correction [13], although conflicting findings persist [12,14]. Several clinical studies have reported SL rates, yet the clinical significance of SL remains ambiguous [12,15]. While some studies have shown a strong correlation between SL and pseudoarthrosis, they found no significant association with clinical pain scores [16,17]. In contrast, other studies have reported notable improvements in Visual Analog Scale (VAS) and Oswestry Disability Index (ODI) scores in patients without SL [11–13]. Analyzing subject-specific factors to predict SL can help improve surgical outcomes by enabling personalized fusion strategies and treatment adjustments [2]. Additionally, anticipating long-term outcomes can inform targeted post-surgical recommendations for at-risk patients [17].
The inconclusive findings regarding the risk of SL following LIF and its clinical significance underscore the need for a more quantitative approach. Given the lack of meta-analyses on risk factors for SL in patients with degenerative lumbar disease, a comprehensive literature review is warranted to systematically evaluate the risk factors for SL. This meta-analysis was conducted with the following aims: (1) identify the risk factors for SL following LIF in patients with degenerative lumbar disease; (2) assess the clinical relevance of SL in terms of patient-reported outcomes; and (3) evaluate the prognostic value of Hounsfield units (HU) in predicting SL.
Methods
Literature search and selection
A comprehensive literature search was conducted across ScienceDirect, PubMed, Google Scholar, Epistemonikos, and Cochrane databases, up to October 2024, without language restrictions. The search utilized Medical Subject Headings (MeSH) and followed the PICOS (Participants, Intervention, Comparator, Outcome, and Study design) framework, as summarized in Table 1. The aim was to identify cohort studies evaluating risk factors for SL following LIF in patients with degenerative lumbar disease. Additionally, the reference lists of relevant studies were manually screened to identify any additional literature. A total of 22 cohort studies were identified and deemed suitable for inclusion in the quantitative analysis.
Study design and inclusion criteria
This meta-analysis adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [18]. The study protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO) database (ID: CRD42024563780) on June 29th, 2024, prior to the commencement of the systematic search [19]. Three authors conducted the literature search, data extraction, and bias assessment, resolving any discrepancies or uncertainties regarding study eligibility through consensus, and involving an additional author as needed. Retrospective or prospective cohort studies addressing the risk of SL following LIF surgery in degenerative lumbar disease were eligible for inclusion.
The inclusion criteria required that studies assessed at least one risk factor for SL, included patients aged ≥18 years, maintained a minimum follow-up duration of 3 months, and explicitly delineated whether the focus was on single or multi-segment interbody fusion. Included studies involved decompression with LIF utilizing traditional trajectory (TT) or cortical bone trajectory (CBT) and were required to have reported the presence of a degenerative lumbar spine disease, such as spondylolisthesis, lumbar stenosis, or lumbar disc herniation, as verified by imaging studies and clinical symptoms. Studies enrolling patients with successful conservative therapy or recurrent deteriorative lumbar spine disorder refractory to medical treatment (including medication, traction therapy, local injections, or physical rehabilitation) were also included. Included studies were required to identify at least one clinical or radiographic factor linked to SL risk, compare SL and non-SL groups, and provide a clear radiographic definition of SL, whether by measurement or imaging characteristic (double halo sign) [2]. A detailed summary of methodological characteristics is provided in Table 2.
The exclusion criteria included case reports, studies involving pediatric populations, tumor-related lesions of the lumbar spine, deep wound infections, and instances of screw malposition or redirection observed in postoperative imaging. Studies lacking adequate details on LIF or screw fixation characteristics were also excluded.
Quality assessment of included studies
The quality of the included studies was assessed using the Risk of Bias in Non-randomized Studies-of Interventions (ROBINS-I) tool. This tool evaluates the risk of bias across seven domains: bias due to confounding, selection of participants, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes, and selection of reported result. Each study was evaluated and scored based on these criteria.
Data extraction
Data pertaining to patient demographics, baseline characteristics, and risk factors were extracted. Key study characteristics assessed included country of origin, sample size, mean age of patients, percentage of female participants, follow-up duration, surgical specifics (single/multi-level infusion status, posterior spine involvement, decompression, trajectory technique, LIF approach), presence of osteoporosis and its diagnostic method, number of patients undergoing revision surgery, and the number of loosened and non-loosened screws. Baseline comorbidities (such as diabetes and hypertension), smoking status, and sex distribution were analyzed to assess comparability between the SL and non-SL groups.
The extracted data were ratio measures (odds ratio [OR] with upper and lower confidence intervals) and continuous measures (mean difference [MD] with standard deviations). The data were derived from various parameters such as patient age, body mass index (BMI), HU values, postoperative Numeric Rating Scale (NRS) scores, Oswestry Disability Index (ODI), and radiographic measurements (sagittal vertical axis [SVA], pelvic incidence [PI], screw diameter and length, Cobb’s angle, thoracic kyphosis [TK], LL, and sacral slope [SS]). A dichotomized scale was used to generate receiver-operating characteristic (ROC) curves, incorporating sensitivity and specificity. Subgroup analyses of HU values were performed based on the trajectory model (dynamic or static), decompression involvement, fusion level (single, multi, or both), fusion approach, and trajectory technique (CBT or TT).
Data synthesis and analysis
Continuous data were presented as mean±standard deviation, and categorical variables as frequency (percentage). Comparisons between studies were facilitated by the incorporation of 95% confidence intervals (CIs). The restricted maximum likelihood (REML) model was used to combine effect estimates, accounting for variations in treatment effects. The Mantel-Haenszel method was employed to assess heterogeneity, guiding model selection. The analysis was conducted using Stata software ver. 18.0 (Stata Corp., College Station, TX, USA), with statistical significance defined as p<0.05 [20]. Heterogeneity among the included studies was assessed using the I2 statistic. In case of significant heterogeneity (I2>50%), a random-effects model was employed; in the absence of significant heterogeneity, a fixed-effects model was applied.
Results
Study selection
The literature search yielded 511 articles. Of these, 189 duplicate publications were removed, leaving 322 articles. Of the 322 articles, 287 were excluded due to the lack of risk factor assessment, unsuitable study design (case report), and use of specific drugs. The remaining 35 articles underwent full-text screening, of which 13 were excluded due to trauma focus, cervical/thoracic involvement, or non-fusion surgery. Finally, 22 studies were included in the meta-analysis (Fig. 1) [6–8,10–12,14,15,21–34].
The characteristics of the included studies are summarized in Table 3. Nine studies had a follow-up duration of ≤12 months [6,8,22,24–26,32–34], while the remaining 13 studies had a follow-up of >12 months [7,10–12,14,15,21,23,27–31]. Fourteen studies included patients with osteoporosis [6–8,11,14,15,21,24–26,29–31,33], with only six studies partially incorporating such patients [14,15,21,24–26]. One study excluded osteoporosis patients [32], while seven other studies lacked sufficient information in this regard [10,12,22,23,27,28,34]. Seven studies focused on single-segment LIF [8,22,25,26,31–33], two studies on multi-segment fusion [14,24], and 13 studies addressed both single and multi-segment LIF [6,7,10–12,15,21,23,27–30,34]. Additionally, 14 studies exclusively focused on the posterior approach [6,7,10,14,15,21,23–26,31–34]. Nine studies included decompression as part of the surgical procedure [7,11,12,14,15,22,25,26,29]. Two studies exclusively utilized the CBT technique [26,31], while the remaining 20 studies employed the TT technique [6–8,10–12,14,15,21–25,27–30,32–34]. Furthermore, four studies focused solely on dynamic spine stabilization [12,14,15,23], whereas 18 studies investigated static stabilization methods [6–8,10,11,21,22,24–34]. The quality assessment using the ROBINS-I tool, detailed in Fig. 2, showed a low overall risk of bias in most included studies.
Study characteristics
The 22 studies included in this meta-analysis were published from 2011 to 2024 and conducted in various countries, including China, Japan, Taiwan, Russia, and France. The characteristics of the study population for each study are summarized in Table 3. The total sample size was 3,689 patients involving 17,772 screws. The mean age was 61.95±9.55 years, and 56% of the patients were female. Overall, 1,770 screws (10%±2%) were loose in 1,089 patients (29%±5%), with 5%±3% of patients undergoing revision surgery. Comorbidities included diabetes in 410 patients (11.11%) and hypertension in 496 patients (13.45%); 316 patients (8,57%) had a smoking status. Multivariate logistic regression showed no significant differences between the SL and non-SL groups in terms of sex (p=0.16), diabetes (p=0.70), hypertension (p=0.94), or smoking status (p=0.98) (Fig. 3). However, the proportional analysis across studies (Table 4) showed significant differences in the proportion of female patients (0.56; 95% CI, 0.51–0.62; p=0.00; I2=94.98%), patients with SL (0.29; 95% CI, 0.24–0.34; p=0.00; I2=90.96%), number of loosened screws (0.10; 95% CI, 0.08–0.11; p=0.00; I2=77.20%), and patients who underwent revision surgery (0.05; 95% CI, 0.02–0.16; p=0.00; I2=95.96%), with notable heterogeneity in all these outcomes.
Clinical factors
Advancing age (MD: 2.79; 95% CI, 1.78–3.79; p=0.00; I2= 49.57%) was identified as a predictor of SL, but ratio measurements suggested age was not a reliable predictor (OR, 1.01; 95% CI, 0.97–1.04; p=0.72; I2=75.31%). Additionally, BMI (MD, −0.17; 95% CI, −0.50 to 0.17; p=0.33; I2=0.00%) was not a significant risk factor for SL (Table 4).
Spinal and sagittal radiographic parameters
Cobb’s angle (MD, 2.42; 95% CI, 0.36–4.49; p=0.02; I2=0.00%) and sacral fixation (OR, 2.59; 95% CI, 1.85–3.62; p=0.00; I2=15.53%) were associated with SL. Conversely, lower values of LL angle (MD, −3.67; 95% CI, −6.33 to −1.01; p=0.01; I2=4.07%), and HU (MD, −25.30; 95% CI, −31.87 to −18.72; p=0.00; I2=48.50%) also showed a strong association with SL. Ratio measurements of HU also corroborated these findings (OR, 0.98; 95% CI, 0.98–0.99; p=0.00; I2=57.89%). Subsequent pooled analysis indicated that lumbosacral fixation (OR, 1.44; 95% CI, 0.67–3.10; p=0.35; I2=77.40%), pelvic incidence–lumbar lordosis (PI–LL) mismatch (MD, −1.89; 95% CI, −4.85 to 0.17; p=0.21; I2=11.01%), PI (MD, −1.84; 95% CI, −4.08 to 0.40; p=0.11; I2=0.00%), SS (MD, −1.00; 95% CI, −3.45 to 1.45; p=0.42; I2=0.00%), SVA (MD, 5.43; 95% CI, −3.88 to 14.74; p=0.25; I2=64.33%), and TK (MD, 0.33; 95% CI, −2.72 to 3.37; p=0.83; I2=0.00%) were not associated with SL (Fig. 4).
Pedicle screw-related factors
Increased fused levels (OR, 2.62; 95% CI, 1.74–3.92; p=0.00; I2=79.33%) and lower screw length (MD, −1.62; 95% CI, −2.78 to −0.45; p=0.01; I2=0.00%) were associated with SL. However, screw diameter (MD, −0.00; 95% CI, −0.04 to 0.03; p=0.95; I2=0.00%) was not identified as an independent risk factor for SL (Fig. 4, Table 4).
Clinical relevance of SL
SL is an established risk factor for impaired quality of life in several aspects (Table 4). Pooled analysis revealed that SL significantly increased the postoperative ODI (MD, 3.34; 95% CI, 0.49–6.20; p=0.02; I2=62.53%) and back pain NRS scores (MD, 0.75; 95% CI, 0.42–1.07; p=0.00; I2=0.00%).
Subgroup analysis of Hounsfield unit
Subgroup analyses revealed several significant findings regarding HU outcomes (Table 5). Decompression involvement significantly increased the risk of SL (MD, −21.75; 95% CI, −50.97 to −7.47; p=0.00; I2=81.46%), whereas the absence of decompression showed no such association. Dynamic trajectory markedly increased the incidence of SL (MD, −9.19; 95% CI, −61.18 to 42.80; p=0.00; I2=92.08%), in contrast to a static trajectory, which did not yield similar results. The use of single, multi, or combined single and multi-segment fusion did not appear to increase the risk of SL. Posterior approach was associated with a significantly higher risk of SL (MD, −25.30; 95% CI, −31.87 to −18.72; p=0.00; I2=80.50%). Additionally, TT was associated with a significant increase in SL incidence (MD, −25.30; 95% CI, −31.87 to −18.72; p=0.05; I2=48.50), whereas the impact of CBT could not be determined due to insufficient studies.
Prognostic capability of HU for predicting SL
The pooled sensitivity and specificity HU in predicting SL were 0.74 (95% CI, 0.67–0.81; p=0.04; I2=49.74%) and 0.76 (95% CI, 0.66–0.84; p=0.00; I2=92.71%), respectively (Fig. 5). Furthermore, the area under the curve (AUC) from ROC curve analysis was 0.80 (95%, 0.77–0.84), indicating a very good prognostic value [35].
Discussion
This meta-analysis provides quantitative evidence on risk factors for SL after LIF in degenerative lumbar disease. To the best of our knowledge, this is the first study to systematically analyze predictive factors for SL and elucidate its clinical importance by incorporating multiple cohorts. Additionally, this is the first study to evaluate the prognostic utility of HU value as a predictor of SL. The findings can help neurosurgeons anticipate and prepare for early intervention in patients at high risk for SL.
The study examined patients’ comorbidities to minimize confounding bias and found no significant differences between the SL and non-SL groups. This is important because several studies have identified that factors such as female sex [36], diabetes [37], hypertension, osteoporosis [38], smoking status [39], chronic kidney disease [38], and chronic steroid use [40] can influence SL. However, the majority of proportion analyses showed significant differences in certain comorbidities (female sex, number of loosened screws, revision surgery, and screw settings), highlighting the need for more optimal settings. Additionally, proportional analysis revealed a significant difference when comparing dynamic and static trajectories; however, analysis conducted solely on the dynamic trajectory did not yield statistically significant results.
It is often hypothesized that SL increases the range of motion in stabilized lumbar spine segments, potentially affecting the patient’s quality of life. However, studies have shown conflicting results in this respect. Ko et al. [41] in 2010 found significant improvements in VAS scores for low-back pain and ODI scores for functional disability, regardless of SL. Similarly, Schaeren et al. [42] in 2003 reported no symptoms or back pain due to screw breakage or loosening. Our findings, however, indicate that SL significantly increases back pain NRS and ODI scores, highlighting the need for better anticipation of SL.
Altered bone mineral density (BMD) has been identified as a predictor of pedicle SL [11]. While dual-energy X-ray absorptiometry (DEXA) is commonly used to assess BMD [43], its accuracy can be affected by osteophyte formation, facet joint degeneration, and aortic calcification in patients with degenerative lumbar diseases.
Measuring radiodensity in HU may offer a convenient alternative for bone quality assessment because not all patients undergo routine DEXA exams before surgery [44]. Increasing evidence suggests that radiodensity measured in HU is strongly linked to BMD and may predict low-energy fractures, implant instability, and pseudoarthrosis [5,45]. Our analysis revealed a positive correlation between pedicle SL rates and decreased radiodensity (HU value). The HU prognostic model showed a high AUC value of 0.80 (0.77–0.84) in our ROC analysis, indicating good predictive value for screw loosening. This outperforms a previous study by Lee et al. [46] in 2022, which reported an AUC of 0.660 (0.541–0.766) incorporating the thoracic spine segment. The findings suggest that HU value is a superior prognostic tool compared to the conventional T-score for assessing SL risk. While this finding underscores the HU score’s significance as a major predictive factor for SL, it is imperative to recognize that bone quality is not the sole contributing factor, as suggested by various other studies [2,47].
In spinal surgery, decompression—typically via laminectomy or foraminotomy—relieves neural compression but may compromise spinal stability, particularly affecting the integrity of pedicle screw fixation. Posterior screw placements also experience directional forces during movement and loading, causing micromotion and potential loosening. Additionally, longer fusion constructs alter stress distribution, increasing the risk of loosening, especially at the cranial and caudal ends of the construct. Yuan et al. [48] in 2025 reported a high rate of SL in patients undergoing posterior decompression and instrumentation, especially with long fusion constructs. Under dynamic loading, micro-movements between the implant and bone can lead to SL, a key concern in dynamic stabilization systems that require both mobility and secure fixation [49]. Dynamic screw trajectories engaging only the cortical bone may induce uneven stress distribution, increasing the risk of loosening and fatigue cracks over time [50]. The TT for pedicle screw fixation targets cancellous bone along the pedicle axis but offers limited anterior column support. Under dynamic loads, TT screws show reduced stability, increased micromotion, and higher risk of loosening. TT screw may also contribute to adjacent segment hypermobility, compromising overall fixation [51]. Although the studies included in this review did not directly compare surgical approaches regarding SL, we analyzed HU unit variations across these settings. The significant differences in HU values among patients undergoing decompression, dynamic trajectories, and posterior approaches emphasize the need for a more precise interpretation in assessing the correlation of HU values with SL incidence.
Our study identified advanced age, but not BMI, as an independent predictor of lumbar SL. Although high BMI appears to be medically relevant due to increased mechanical stress, other factors may contribute more significantly toward SL [52]. The identification of advanced age as a predictor of SL aligns with other research, such as Kim et al. [47] in 2015, which found patients over 65 with S1 pedicle screw fixation were more prone to SL. Decreased bone density with advancing age may contribute to this increased risk [53]. The analysis found a significant correlation between age and SL when age was measured as continuous data, but not when using ratio measurements. Further research is needed to clarify these findings due to limited information and potential confounding biases in existing studies.
While the correlation between Cobb angle and SL is not well-established, our pooled analysis suggests that Cobb angle is an independent predictor of lumbar SL. The Cobb angle measures spinal curvature severity, particularly in scoliosis [54]. A higher Cobb angle, indicating more severe curvature, can lead to uneven load distribution on screws. Moreover, it often necessitates complex surgeries with longer fusion segments and extensive instrumentation, increasing the risk of loosening [38]. Further research is warranted to confirm these findings, as this topic has not been specifically studied before. Our findings also support sacral fixation and increased fused levels as predictors of lumbar SL. A study by Tokuhashi et al. [16] in 2008, which included osteoporotic patients, reported SL rate of 7.4% for 1-level lumbar fixation and over 40% for four levels of fixation at ≥3 years follow-up. Fusion to S1 significantly contributes to SL, with reported rates ranging from 15.6% to 46.5% [55]. The sacrum’s anatomy, with its larger pedicle diameter and shorter length, results in less cortical fixation at the screw-bone interface. Additionally, the sacrum is primarily composed of cancellous bone, which may explain the higher rate of SL in S1 [56]. PI, a key spinal pelvic parameter influencing spinal sagittal balance, is determined by the anatomical position of the pelvis [57]. PI exhibits a strong correlation with LL and SS angle; a lower PI indicates flattened lordosis and more vertical sacrum, which worsens sagittal balance and decreases biomechanical instability, increasing the risk of SL [58]. Kim et al. [47] in 2015 only identified high PI and PI–LL mismatch (>10°) as independent predictors of S1 pedicle SL, but not high SS or low LL. Interestingly, our analysis identified a lower LL angle as a significant predictor of lumbar SL, but not high PI, SS, or PI–LL mismatch, indicating the need for further research. SVA and TK are also important parameters for assessing sagittal malalignment. SVA imbalance and abnormal TK, potentially due to disrupted LL compensation, can alter force distribution along the spine, impacting screw stability [59]. However, our analysis did not identify SVA and TK as significant predictors of SL, indicating the need for further research.
Our study identified screw length as an independent predictor of SL, but not screw diameter. Computational model analysis by Matsukawa et al. [60] in 2020 suggests that larger diameters and longer screws can provide better fixation by engaging more bone, enhancing stability and reducing loosening risk. Screw diameter impacts resistance to pullout and flexion-extension loading, while screw depth affects resistance to lateral bending and axial rotation [60]. Our conflicting results regarding screw diameter may be due to the limited number of studies and inadequate screw characteristic data. Other factors, such as anterior cortex penetration or patient vertebral morphology, might play a more significant role [61,62].
Some limitations of our study should be acknowledged. The variability in follow-up durations (3–43.4 months) can impact outcomes and hinder a comprehensive evaluation of predictive factors and clinical outcomes. Shorter follow-up periods suggest that SL occurs earlier than detected by imaging due to clinical silence. The true impact on clinical outcomes can only be assessed with longer follow-up. This variability may also contribute to heterogeneity in several outcomes. The study’s results may have been influenced by the broad spectrum of patient characteristics and varying measurement protocols, potentially introducing bias. Additionally, the lack of a clear or consistent cut-off value for ratio measurements in binary groups can complicate data synthesis and interpretation. Lastly, there is a paucity of studies focusing on certain parameters, particularly sagittal radiographic parameters and screw characteristics, despite the significance of such comparison. This may explain the conflicting outcomes compared to several previous studies.
Conclusions
This study identified significant risk factors for SL following LIF in patients with degenerative lumbar disease. SL was found to be a clinically significant complication, and HU showed robust prognostic power for predicting loosened screws with high sensitivity and specificity. Our findings indicate that advancing age, Cobb angle, sacral fixation, increased fused levels, lower LL angle, screw length, and HU units are significantly associated with SL. Subgroup analyses revealed additional factors correlated with SL, including decompression, dynamic trajectory, posterior approach, and TT. SL was also linked to increased back pain and functional outcomes. These findings advance understanding of SL management in neurosurgery, emphasizing the importance of recognizing risk factors in post-LIF care of patients with degenerative lumbar diseases. This knowledge can lead to more precise and effective treatment strategies.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Full text via PMC
Download Citation
