Vertebral infection after kyphoplasty: distinguishing misdiagnosis from true postoperative infection and implications for management

Abstract

Background

Spinal infection following percutaneous kyphoplasty (PKP) is rare but clinically devastating. However, its true incidence, diagnostic challenges, and optimal revision strategies remain poorly defined, particularly in referral-based cohorts.

Methods

We conducted a retrospective study comprising two components: (1) an institutional incidence analysis of 6,328 PKP procedures performed between 2015 and 2025, and (2) a pooled clinical cohort of 39 patients treated for post-PKP spinal infection, including 3 hospital-acquired cases and 36 referred cases. Patients were managed with either surgical (n = 27) or conservative (n = 12) strategies. Within the surgical group, a selective cement management approach (complete removal vs. partial retention) was applied based on intraoperative assessment. Clinical, microbiological, and radiographic data were analyzed descriptively. Functional outcomes were assessed using the Oswestry Disability Index (ODI).

Results

The institutional incidence of post-PKP infection was 0.05% (3/6328). In the pooled cohort (n = 39), the median age was 72 years, all patients presented with back pain, and fever was present in 20 patients. Potential etiologies included hematogenous spread, preoperative misdiagnosis, and iatrogenic factors. Eight cases of spinal tuberculosis were initially misdiagnosed as osteoporotic fractures. Microbiological confirmation was limited, with low culture positivity. Surgical management was associated with lower ODI scores at final follow-up compared to conservative treatment; however, interpretation is limited by baseline imbalance and selection bias. Among surgically treated survivors, infection control and radiographic fusion were observed at final follow-up. No recurrence was recorded in 8 carefully selected patients managed with partial cement retention; however, this observation remains limited by the small subgroup size and the absence of matched controls.

Conclusions

Post-PKP vertebral infection is uncommon but clinically challenging and often complicated by diagnostic uncertainty. Surgical debridement and stabilization may be useful in selected patients. Partial cement retention can be considered only in selected situations, and its wider use requires further validation.

Introduction

With the increase in the age of the population, the incidence rate of osteoporosis increases yearly. Several older patients with osteoporosis are experiencing thoracolumbar fractures, leading to back pain, severely affecting their daily lives [1, 2]. Vertebral augmentation was initially introduced as percutaneous vertebroplasty in 1987 by Galibert et al. for spinal hemangioma, whereas kyphoplasty was developed later in the late 1990s [3]. Kyphoplasty has been widely used in the treatment of osteoporotic vertebral compression fractures, vertebral osteolytic metastases, vertebral hemangiomas, and other diseases [4–6]. With the increasing application of PKP, accompanying complications, including bone cement leakage, nerve compression, re-fracture, infection, and bone cement embolism, have also significantly increased [7–9].

Deep infections after vertebral augmentation surgery are rare, and previous studies have generally been limited by small sample sizes. To date, one of the largest studies investigating spinal infections following vertebral augmentation included only 33 patients [10], with a reported rate of 0.32% (18/5,749) [11]. More recently, a systematic review in Global Spine Journal further confirmed that infection following vertebroplasty or kyphoplasty is uncommon but clinically serious, with potentially severe outcomes requiring individualized treatment strategies [12]. In addition, diagnostic uncertainty remains an important challenge. A recent meta-epidemiological study in The Spine Journal demonstrated wide variability in the definitions used for native vertebral osteomyelitis, highlighting the lack of a unified diagnostic framework [13]. This diagnostic heterogeneity may contribute to delayed recognition or misdiagnosis of primary vertebral infection as osteoporotic vertebral compression fracture before PKP. The current treatment methods for spinal infections after vertebral augmentation surgery include conservative antibiotic therapy, simple anterior surgery, simple posterior surgery, and combined anterior posterior approach surgery. Presently, studies on spinal infection after PKP surgery generally focus on the etiology, risk factor analysis, pathogen distribution, indications, and methods of conservative and surgical treatment. The study sample rarely exceeded 30 patients, and no study has reported exploring the effect of bone cement removal and retention in infected vertebral bodies on healing. In a posterior surgery for spinal infection after PKP, complete removal of bone cement requires a longer surgical time, more bleeding, higher risk of nerve damage, and anesthesia risk. Therefore, the present study aimed to clarify the clinical characteristics, diagnostic pitfalls, presumed infection pathways, and treatment outcomes of vertebral infection after PKP, with particular attention to individualized revision strategies and selective cement management.

Patients and methods

Study design, data collection, and cohort definition

This retrospective study consisted of two analytically distinct components:

1. Institutional incidence cohort: All patients (6,328) who underwent PKP at our institution between January 2015 and June 2025 were reviewed to determine the incidence of post-procedural spinal infection.

2. Clinical management cohort: A pooled cohort of 39 patients diagnosed with spinal infection following PKP was analyzed for clinical characteristics and treatment outcomes. This cohort included 3 hospital-acquired cases and 36 patients referred from external institutions (Fig. 1)

Fig. 1.

Flowchart of cohort assembly and study design. Legend: Flow diagram illustrating the assembly of the study cohorts and the two-component study design. From January 2015 to June 2025, 6,328 patients underwent percutaneous kyphoplasty (PKP) at our institution. Among these patients, 3 hospital-acquired cases of post-PKP vertebral infection were identified and used for the institutional incidence analysis. In addition, 36 patients with vertebral infection after PKP were referred from external institutions. These 3 institutional cases and 36 referral cases together formed the pooled clinical management cohort (n = 39), which was analyzed for clinical characteristics, treatment strategies, and outcomes. Within the pooled cohort, 27 patients underwent surgical treatment and 12 received conservative treatment

For the institutional incidence cohort, we reviewed 6,328 PKP procedures performed at our institution between January 2015 and June 2025. For the pooled clinical management cohort, we collected demographic, laboratory, imaging, microbiological, treatment, and follow-up data from 39 patients diagnosed with vertebral infection after PKP. Follow-up ranged from 8 months to 10 years. The cohort assembly and two-component study design are shown in Fig. 1. This study was approved by the Ethics Committee (Ethical approval No. (204–013))22) YYL-K (2026496) and complies with all regulations. In accordance with the ethical principles outlined in the Declaration of Helsinki, all participants provided informed consent before participating in the study. The anonymity and confidentiality of the participants were guaranteed, and participation was completely voluntary. Clinical trial number: not applicable.

Inclusion criteria and exclusion criteria

Inclusion criteria

Patients were included if they met at least two of the following criteria:

• ① Clinical signs suggestive of infection (fever, persistent pain, local tenderness)

• ② Elevated inflammatory markers (CRP, ESR, or WBC)

• ③ Radiological evidence of infection (MRI or CT findings consistent with spondylodiscitis or abscess)

• ④ Microbiological or histopathological confirmation when available

Exclusion criteria

• ① Non-infectious causes of vertebral collapse

• ② Incomplete follow-up (< 6 months)

• ③ Malignant spinal tumors with secondary infection

Patients with prior antibiotic exposure, prior external biopsy, paravertebral or psoas abscess, or previous external treatment were not automatically excluded if the infection was considered related to a prior PKP procedure and sufficient follow-up data were available.

Definition of infection and clinical outcomes

Infections were classified into three levels according to diagnostic certainty:

• Confirmed infection: Definitive evidence from microbiology or histopathology

• Probable infection: Consistent clinical, laboratory, and radiological findings

• Possible infection: Suggestive but non-definitive findings

Outcomes were clearly defined and reported separately:

• Infection control: Resolution of clinical symptoms and normalized inflammatory markers

• Radiographic bony fusion: Evaluated by postoperative CT imaging

• Functional outcome: Assessed using the Oswestry Disability Index (ODI)

• Recurrence-free survival: No evidence of infection relapse during follow-up

Treatment strategies

Patients were managed with either:

• Surgical treatment: posterior debridement, stabilization, and bone grafting

• Conservative treatment: prolonged antibiotic therapy

The decision for surgery was based on clinical condition, neurological status, spinal stability, and patient tolerance.

Cement management strategy

Within the surgical group, the decision to remove or retain bone cement was determined intraoperatively by the surgical team based on:

• Degree of cement integration with surrounding bone

• Presence of vertebral wall destruction

• Epidural involvement or neural compression

• Estimated risk of destabilization or excessive bleeding

This decision was surgeon-driven and predefined, rather than patient preference (Fig. 2).

Fig. 2.

Strategic Surgical Decision-making Algorithm for Post-PKP Spinal Infection. Legend: The algorithm illustrates the structured clinical decision-making pathway for managing spinal infections after percutaneous kyphoplasty (PKP)

• Triage Phase: Following diagnosis (N = 39), patients were triaged into Surgical (n = 27) or Conservative (n = 12) groups based on physical status and surgical tolerance.

• Intraoperative Decision Phase: Within the surgical cohort, the choice between Complete Cement Removal (n = 19) and Partial Cement Retention (n = 8) was determined by the degree of cement integration with the vertebral wall.

• Surgical principle in selected patients: Both surgical subgroups underwent thorough debridement to disrupt bacterial biofilms and received localized antibiotic carriers.

• Final Clinical Outcome: The combination of strategic surgical intervention and prolonged systemic antibiotic therapy was associated with infection control and radiographic fusion in surviving surgically treated patients.

Microbiological analysis

Specimens included blood, intraoperative tissue, and paraspinal samples when available. Culture methods, histopathology, and tuberculosis-specific testing were performed according to institutional protocols. The isolation media for the drug susceptibility tests were 90-mm Chinese blue agar medium (Shanghai Kemajia Biotechnology Co., Ltd. Shanghai, China) and Colombian blood agar medium (Shanghai Kemajia Biotechnology Co., Ltd. Shanghai, China). Pathogen identification and drug susceptibility test were performed using the VITEK-2 Compact automatic bacterial tester (BioMerieux, Hazelwood, MO, USA). Drug sensitivity was evaluated according to the Clinical and Laboratory Standards Institute (CLSI, 2017) standards. Due to prior antibiotic exposure in many referral cases, culture yield was limited.

Statistical analysis

All statistical analyses were performed using SPSS version 25.0 (IBM Corp., Armonk, NY, USA). Given the small sample size and heterogeneity of the cohort, analyses were primarily descriptive. Continuous variables were assessed for normality and are presented as mean ± standard deviation or median (interquartile range), as appropriate. Categorical variables are reported as counts and percentages.

For between-group comparisons, nonparametric methods were preferentially applied due to the limited sample size. Continuous variables were compared using the Mann–Whitney U test, and categorical variables were analyzed using the chi-square test or Fisher’s exact test, as appropriate. All statistical tests were two-sided, and a P value < 0.05 was considered statistically significant. Given the small sample size and heterogeneity of the cohort, results were mainly summarized using medians, interquartile ranges, counts, and percentages. Comparative analyses were considered exploratory because of the retrospective design and potential selection bias.

Results

Cohort characteristics

A total of 39 patients with vertebral infection after percutaneous kyphoplasty were included in this retrospective cohort. Among them, 3 cases occurred after PKP performed at our institution, whereas 36 patients were referred from external hospitals for further management. Twenty-seven patients underwent surgical treatment, and 12 patients received conservative treatment.

The median age of the overall cohort was 72 years (IQR, 69–77), and 15 patients were male (38.5%). Comorbidities included hypertension in 14 patients (35.9%), diabetes mellitus in 2 patients (5.1%), coronary artery disease in 3 patients (7.7%), chronic kidney disease in 2 patients (5.1%), and steroid or immunosuppressive use in 4 patients (10.3%). The median pre-treatment CRP level was 20.4 mg/L (IQR, 11.6–58.0), and the median ESR was 50 mm/h (IQR, 24.5–78.5). Fever was present in 20 patients (51.3%), and all patients presented with back pain. Preoperative neurological impairment, defined as ASIA grade D, was identified in 10 patients (25.6%). Eight patients (20.5%) were diagnosed with spinal tuberculosis, and 31 patients (79.5%) were classified as having non-specific spinal infection (Table 1).

Table 1.

Baseline characteristics of the overall cohort

Variable	Total cohort (n = 39)
Age, years	72 (69–77)
Male sex	15 (38.5%)
Referral cases	36 (92.3%)
Institutional cases	3 (7.7%)
Diabetes mellitus	2 (5.1%)
Hypertension	14 (35.9%)
Coronary artery disease	3 (7.7%)
Chronic kidney disease	2 (5.1%)
Steroid or immunosuppressive use	4 (10.3%)
Serum albumin, g/L	33.2 (31.9–36.8)
WBC, ×10⁹/L	6.2 (4.5–8.1)
CRP, mg/L	20.4 (11.6–58.0)
ESR, mm/h	50 (24.5–78.5)
Fever	20 (51.3%)
Back pain	39 (100.0%)
Preoperative neurological impairment (ASIA grade D)	10 (25.6%)
Spinal tuberculosis	8 (20.5%)
Non-specific spinal infection	31 (79.5%)
Surgical treatment	27 (69.2%)
Conservative treatment	12 (30.8%)
Complete cement removal	17 (43.6%)
Partial cement retention	8 (20.5%)
Posterior fixation only	2 (5.1%)
Preoperative ODI score	40 (25–55)
Final ODI score	13 (7–18)
Mortality	3 (7.7%)

Values are presented as median (interquartile range) or n (%)

ASIA American Spinal Injury Association, CRP C-reactive protein, ESR Erythrocyte sedimentation rate, ODI Oswestry Disability Index, WBC White blood cell count

Complete pre-PKP clinical and imaging records were available in 20 of 39 patients. The remaining 19 patients had incomplete preoperative records because they were referred from external institutions. Therefore, analyses of presumed preoperative misdiagnosis and remote infectious foci were considered descriptive and exploratory.

Infection pathways, pathogen categories, and clinical flow

The distribution of presumed infection pathways, pathogen categories, treatment strategies, and outcomes is summarized in Fig. 3. Among the 39 patients, the most common presumed infection pathway was hematogenous spread (n = 20), followed by preoperative misdiagnosis (n = 14) and iatrogenic contamination (n = 5). Of the hematogenous cases, 4 were categorized as spinal tuberculosis and 16 as non-specific spinal infection. Of the preoperative misdiagnosis cases, 2 were categorized as spinal tuberculosis and 12 as non-specific spinal infection, whereas all 5 iatrogenic contamination cases were categorized as non-specific spinal infection.

Fig. 3.

Flow from infection pathway to pathogen category, treatment, and outcome. Flow counts were confirmed from the final dataset: hematogenous spread, spinal tuberculosis 4 and non-specific spinal infection 16; preoperative misdiagnosis, spinal tuberculosis 2 and non-specific spinal infection 12; iatrogenic contamination, non-specific spinal infection 5. Spinal tuberculosis cases: surgical 6, conservative 2. Non-specific spinal infection cases: surgical 21, conservative 10. Surgical treatment outcomes: infection control 25, death 2. Conservative treatment outcomes: infection control 11, death 1

Overall, 8 patients were classified as having spinal tuberculosis and 31 as having non-specific spinal infection. In the tuberculosis subgroup, 6 patients underwent surgical treatment and 2 received conservative treatment. In the non-specific spinal infection subgroup, 21 patients underwent surgical treatment and 10 received conservative treatment. At the final follow-up, 25 of 27 surgically treated patients and 11 of 12 conservatively treated patients achieved infection control, whereas 2 surgical patients and 1 conservatively treated patient died.

Figure 3 provides a visual summary of the clinical flow of the cohort and highlights that not all cases represented true de novo postoperative infection. In particular, the preoperative misdiagnosis subgroup supports the observation that some patients had pre-existing vertebral infection that had initially been mistaken for osteoporotic vertebral compression fracture before kyphoplasty (Fig. 3).

Comparison between surgical and conservative treatment groups

Baseline characteristics of the surgical and conservative groups are summarized in Table 2. The surgical group included 27 patients, and the conservative group included 12 patients. The surgical group tended to be younger than the conservative group, although the difference did not reach statistical significance [71 years (IQR, 68.5–72.5) vs. 77 years (IQR, 70.5–80.5), P = 0.077] (Table 2).

Table 2.

Comparison between surgical and conservative treatment groups

Variable	Surgical group (n = 27)	Conservative group (n = 12)	P value
Age, years	71 (68.5–72.5)	77 (70.5–80.5)	0.077
Male sex	9 (33.3%)	6 (50.0%)	0.478
Referral cases	25 (92.6%)	11 (91.7%)	1.000
Diabetes mellitus	2 (7.4%)	0 (0.0%)	1.000
Hypertension	10 (37.0%)	4 (33.3%)	1.000
Coronary artery disease	1 (3.7%)	2 (16.7%)	0.219
Chronic kidney disease	1 (3.7%)	1 (8.3%)	0.526
Steroid or immunosuppressive use	2 (7.4%)	2 (16.7%)	0.573
Serum albumin, g/L	32.5 (31.9–36.5)	34.3 (32–37.2)	0.643
WBC, ×10⁹/L	6.2 (4.6–7.9)	6.0 (4.4–8.2)	0.867
CRP, mg/L	15 (11.4–58.0)	26.8 (16.5–43.0)	0.403
ESR, mm/h	44 (24.5–74)	66 (28.8–84.5)	0.369
Fever	12 (44.4%)	8 (66.7%)	0.301
Back pain	27 (100.0%)	12 (100.0%)	1.000
Preoperative neurological impairment (ASIA grade D)	7 (25.9%)	3 (25.0%)	1.000
Spinal tuberculosis	6 (22.2%)	2 (16.7%)	1.000
Preoperative ODI score	40 (22.5–57.5)	35 (28.8–40)	0.375
Final ODI score	9 (6.5–14)	27.5 (18.2–35)	< 0.001
Mortality	2 (7.4%)	1 (8.3%)	1.000

Values are presented as median (interquartile range) or n (%). Continuous variables were compared using the Mann–Whitney U test. Categorical variables were compared using Fisher’s exact test. P values are provided for descriptive comparison only

There were no statistically significant differences between the two groups in sex distribution, referral status, diabetes mellitus, hypertension, coronary artery disease, chronic kidney disease, steroid or immunosuppressive use, albumin level, WBC count, CRP level, ESR level, fever, preoperative neurological status, tuberculosis proportion, or preoperative ODI score.

The median preoperative ODI was 40 (IQR, 22.5–57.5) in the surgical group and 35 (IQR, 28.8–40) in the conservative group (P = 0.375). At the final follow-up, the surgical group had a lower ODI score than the conservative group [9 (IQR, 6.5–14) vs. 27.5 (IQR, 18.2–35), P < 0.001] (Fig. 4).

Fig. 4.

Comparison of final Oswestry Disability Index scores between surgical and conservative treatment groups. Legend: Box-and-whisker plot comparing final Oswestry Disability Index (ODI) scores between the surgical and conservative groups. Boxes represent the interquartile range, central lines represent medians, and whiskers represent the data range. The comparison was descriptive because treatment allocation was not randomized

This comparison was descriptive because treatment allocation was not randomized and may have been influenced by patient condition, surgical tolerance, infection severity, and physician judgment.

Institutional and referral cases

Among the 39 patients in the pooled clinical management cohort, 3 were institutional cases and 36 were referral cases. Accordingly, the institutional incidence was calculated using only the 3 hospital-acquired cases as the numerator and the 6,328 institutional PKP procedures as the denominator. Because referral cases accounted for most of the pooled cohort, the 39-patient series should be interpreted as a referral-enriched clinical management cohort rather than as a denominator-defined incidence cohort. This distinction avoids denominator bias and prevents overestimation of the institutional infection rate. The three institutional infections occurred sporadically during the 10-year study period, and no temporal clustering was observed. Because only three institutional infections were identified, formal temporal trend analysis was not feasible.

Surgical treatment and cement management

Among the 27 surgically treated patients, 17 underwent complete cement removal, 8 underwent partial cement retention, and 2 underwent posterior fixation without cement removal or extensive lesion debridement (Figs. 5 and 6).

Fig. 5.

Representative case of posterior debridement with partial bone cement retention. Typical case of posterior debridement with partial bone cement retention. Legend: Radiographic timeline of a selected patient managed with partial cement retention. Preoperative MRI and CT showed inflammatory signal changes, prevertebral abscess, and bone destruction around the previous cement. Immediate postoperative radiographs confirmed stable instrumentation with intentionally retained bone cement. Twelve-month follow-up CT and MRI demonstrated radiographic fusion and resolution of active inflammatory findings

Fig. 6.

Representative case of posterior debridement with complete bone cement removal. Typical case of posterior debridement with complete bone cement removal. Legend: Imaging of a patient managed with thorough debridement and complete removal of the bone cement. Preoperative images show extensive vertebral collapse. Postoperative images demonstrate reconstruction with a titanium mesh cage and bone graft, achieving stable fusion and infection control at follow-up

The comparison between complete cement removal and partial cement retention is shown in Table 3. Patients in the cement-retention subgroup had a significantly shorter operative time than those in the cement-removal subgroup [222.5 min (IQR, 202.5–250) vs. 305 min (IQR, 290–360), P < 0.001].

Table 3.

Comparison between complete cement removal and partial cement retention among surgically treated patients

Variable	Complete cement removal (n = 17)	Partial cement retention (n = 8)	P value
Age, years	72 (69–72)	68.5 (65.5–70.2)	0.094
Male sex	4 (23.5%)	4 (50.0%)	0.359
Referral cases	16 (94.1%)	7 (87.5%)	1.000
Diabetes mellitus	1 (5.9%)	0 (0.0%)	1.000
Hypertension	7 (41.2%)	3 (37.5%)	1.000
Serum albumin, g/L	32.5 (31.7–35.8)	32.8 (32.2–36.5)	0.605
WBC, ×10⁹/L	7.2 (4.9–8.6)	4.9 (4.2–6.2)	0.044
CRP, mg/L	22.3 (11.8–73.9)	12.4 (4.9–14.7)	0.232
ESR, mm/h	56 (37–82)	24.5 (14–50.2)	0.137
Fever	7 (41.2%)	4 (50.0%)	1.000
Preoperative neurological impairment (ASIA grade D)	5 (29.4%)	1 (12.5%)	0.624
Spinal tuberculosis	3 (17.6%)	2 (25.0%)	1.000
Preoperative ODI score	40 (10–60)	45 (35–57.5)	0.396
Operative time, min	305 (290–360)	222.5 (202.5–250)	< 0.001
Intraoperative blood loss, mL	1000 (700–1000)	700 (600–825)	0.224
Final ODI score	9 (7–14)	7 (4.8–7.5)	0.134
Mortality	1 (5.9%)	1 (12.5%)	1.000

Values are presented as median (interquartile range) or n (%). The two patients who underwent posterior fixation only were not included in this subgroup comparison. Continuous variables were compared using the Mann–Whitney U test. Categorical variables were compared using Fisher’s exact test

Intraoperative blood loss was numerically lower in the cement-retention subgroup [700 mL (IQR, 600–825)] than in the cement-removal subgroup [1000 mL (IQR, 700–1000)], although this difference was not statistically significant (P = 0.224).

There were no statistically significant differences between the cement-removal and cement-retention subgroups in age, sex, referral status, diabetes mellitus, hypertension, albumin level, fever, preoperative neurological status, CRP, ESR, tuberculosis proportion, preoperative ODI, final ODI, or mortality (Table 3).

Antibiotic treatment and microbiological findings

Antibiotic therapy was administered according to the suspected or confirmed infection type, culture results when available, pathological findings, and clinical response. The diagnosis of spinal tuberculosis was based on histopathological evidence of granulomatous inflammation and/or caseous necrosis, together with compatible clinical and radiological findings; acid-fast staining, TB-PCR, or mycobacterial culture was performed when available but was not consistently available in all retrospective referral cases. Among the 8 patients diagnosed with spinal tuberculosis, standard quadruple anti-tuberculosis therapy consisting of isoniazid, rifampicin, pyrazinamide, and ethambutol was administered during hospitalization. After discharge, oral anti-tuberculosis therapy was continued for 12–18 months according to clinical symptoms, inflammatory markers, imaging findings, and drug tolerance.

Among the 31 patients classified as having non-specific spinal infection, broad-spectrum antibiotic therapy was initially used. The most frequently used regimens included cefoperazone-sulbactam, piperacillin-tazobactam, carbapenems, moxifloxacin, and amikacin. Combination therapy with vancomycin was used in 14 patients, and linezolid-containing regimens were used in 8 patients. Antibiotic regimens were adjusted according to culture results, pathological findings, inflammatory markers, and clinical response whenever possible (Table 4).

Table 4.

Antibiotic treatment and microbiological findings

Variable	Total cohort (n = 39)
Infection type
Spinal tuberculosis	8 (20.5%)
Non-specific spinal infection	31 (79.5%)
Anti-tuberculosis treatment
Quadruple anti-tuberculosis therapy during hospitalization	8 (20.5%)
Post-discharge oral anti-tuberculosis therapy for 12–18 months	8 (20.5%)
Antibiotic treatment for non-specific spinal infection
Cefoperazone-sulbactam-based regimen	20/31 (64.5%)
Piperacillin-tazobactam-based regimen	4/31 (12.9%)
Carbapenem-based regimen	4/31 (12.9%)
Moxifloxacin-based regimen	2/31 (6.5%)
Amikacin-based regimen	1/31 (3.2%)
Vancomycin-containing regimen	14/31 (45.2%)
Linezolid-containing regimen	8/31 (25.8%)
Microbiological findings
Positive blood culture	2/39 (5.1%)
Positive tissue or pus culture	2/39 (5.1%)
Any positive culture	4/39 (10.3%)
Culture-negative cases	35/39 (89.7%)
Identified pathogens	Salmonella spp.; Escherichia coli; Streptococcus gordonii

Values are presented as n (%) or n/N (%). Percentages for antibiotic regimens for non-specific spinal infection were calculated using the 31 patients with non-specific infection as the denominator. Percentages for microbiological findings were calculated using the total cohort as the denominator. Some patients received combination therapy; therefore, antibiotic categories are not mutually exclusive

Positive microbiological results were obtained in 4 patients, including 2 positive blood cultures and 2 positive tissue or pus cultures. The identified pathogens included Salmonella, Escherichia coli, and Streptococcus gordonii. Due to the low culture positivity rate and the high proportion of referred patients with possible prior antibiotic exposure, many cases were managed as culture-negative infections.

For surgically treated patients with non-specific spinal infection, intravenous antibiotics were administered for at least 6 weeks, followed by oral antibiotics for no less than 3 months when clinically indicated. Antibiotic discontinuation was based on clinical improvement and serial normalization of inflammatory markers, including WBC count, CRP, and ESR.

Complications and neurological outcomes

Neurological status was evaluated using the ASIA grading system. Preoperative neurological impairment was identified in 10 patients, all of whom were classified as ASIA grade D. No postoperative deterioration of neurological function was observed in the surgical cohort. Among patients with preoperative neurological impairment, 2 patients showed postoperative improvement in neurological symptoms during follow-up.

Three patients died during follow-up, including 2 patients in the surgical group and 1 patient in the conservative group. The mortality rates were 7.4% in the surgical group and 8.3% in the conservative group, with no statistically significant difference between groups. Because of the limited number of death events, this comparison was considered descriptive and underpowered.

At the final follow-up, no displacement of internal fixation, screw loosening, or rod breakage was observed on plain radiographs in surgically treated survivors. CT imaging demonstrated satisfactory bone fusion in the grafted area, with no recurrent vertebral destruction or sequestrum formation in the originally infected vertebrae. MRI, when performed, showed no progressive inflammatory signal in the infected vertebral body or intervertebral space, and no residual or recurrent paravertebral or psoas abscess.

No recurrence of infection was recorded among surviving patients during the follow-up period. The retrospective design and variable follow-up duration limited the reliability of recurrence-free outcome assessment.

Clinical outcomes

At final follow-up, the surgical group had a lower ODI score than the conservative group. This difference was descriptive, as treatment was selected according to surgical tolerance, infection severity, and physician judgment rather than random allocation. In the cement-retention subgroup, final ODI scores were not worse than those in the cement-removal subgroup [7 (IQR, 4.8–7.5) vs. 9 (IQR, 7–14), P = 0.134]. Because only 8 patients underwent partial cement retention, no firm conclusion regarding equivalence or superiority can be drawn.

Discussion

Principal findings

In this 10-year retrospective study, we analyzed vertebral infection after percutaneous kyphoplasty (PKP) using two analytically distinct perspectives: an institutional incidence cohort and a pooled clinical management cohort enriched by referral cases. The present study yielded several clinically relevant findings. First, the institutional incidence of vertebral infection after PKP was low, which is consistent with previous reports showing that infection after vertebral augmentation is uncommon but clinically serious [10, 14–16]. Second, diagnosis was often challenging in the pooled cohort. Some patients probably had an underlying vertebral infection, including spinal tuberculosis, that was initially interpreted as an osteoporotic compression fracture before cement augmentation. Third, surgical treatment was associated with lower final ODI scores than conservative treatment, although this comparison remains exploratory because treatment allocation was not randomized. In selected patients, partial cement retention did not show an obvious signal of worse functional outcome or recurrence. This finding suggests possible feasibility, but it is not sufficient to establish cement retention as a standard treatment strategy. These findings should be viewed in the context of the cohort structure: only 3 infections occurred after PKP performed at our institution, whereas 36 patients were referred from other hospitals. Therefore, the incidence estimate was based only on institutional cases, while the 39-patient cohort was used to describe clinical management and outcomes.

Incidence, referral bias, and preoperative misdiagnosis

The low institutional incidence observed in the current study is compatible with the literature, which reports that deep infection after vertebral augmentation is rare but potentially devastating [10, 14, 15]. However, interpreting incidence in this disease entity is difficult when referral cases are included. Referred patients are more likely to represent severe, delayed, atypical, or treatment-resistant infections; therefore, they may overrepresent complicated cases and cannot be used to calculate institutional incidence. For this reason, our incidence estimate was restricted to the 3 hospital-acquired cases, while the remaining 36 referral patients were analyzed only for clinical characteristics and treatment outcomes.

One clinically important point from this series is that infection diagnosed after PKP was not always a newly acquired postoperative event. Some patients may have had pre-existing vertebral infection that was misdiagnosed as an osteoporotic compression fracture before cement injection. Previous studies have also emphasized that spinal tuberculosis after vertebral augmentation may represent misdiagnosis rather than a new postoperative complication [17, 18]. Similarly, pyogenic spondylitis or occult infection from other sites may be present before vertebral augmentation and become clinically apparent after the procedure [19, 20]. Reported risk factors such as diabetes, rheumatic disease, hypoproteinemia, longer hospitalization, and pre-existing infection elsewhere may further increase vulnerability to postoperative infectious complications [21–24].

In the current cohort, elevated inflammatory markers before PKP should be interpreted as evidence suggesting possible pre-existing infection rather than as a direct cause of infection. Therefore, the original statement that a high preoperative inflammatory index was the most common cause of post-PKP infection has been revised. A more appropriate interpretation is that unexplained elevation of CRP or ESR, atypical imaging findings, fever, pulmonary or urinary infection, and disproportionate pain should raise suspicion for underlying vertebral infection before PKP. When imaging cannot reliably distinguish infection from osteoporotic fracture, biopsy or additional diagnostic testing should be considered before cement augmentation, provided that the patient’s condition permits.

Some authors have described structural changes after cement augmentation, such as endplate collapse, intervertebral space narrowing, vertebral destruction, and secondary instability [25]. These changes may contribute to pain, mechanical failure, or diagnostic confusion. However, based on the current data, such structural changes should not be presented as the primary mechanism of infection after PKP. The present study supports a more cautious interpretation: post-augmentation infection may arise from true iatrogenic contamination, hematogenous seeding, or missed pre-existing infection, and these mechanisms may coexist in a referral-enriched cohort.

Microbiological findings and antibiotic treatment

Microbiological confirmation was limited in the present study. Only 4 patients had positive cultures, including 2 positive blood cultures and 2 positive tissue or pus cultures. The identified pathogens included Salmonella spp., Escherichia coli, and Streptococcus gordonii. Eight patients were diagnosed with spinal tuberculosis, whereas 31 were classified as non-specific spinal infection. In comparison, Yuan et al. reported Mycobacterium tuberculosis as the most common organism in spinal infection after vertebral augmentation, followed by Staphylococcus aureus, Escherichia coli, and other organisms [14]. The low culture positivity rate was an important limitation and does not indicate absence of infection.

Several factors may explain the low culture yield. Many patients were referred from other hospitals and may have received antibiotics before admission, which could reduce culture positivity. In addition, cement-filled vertebrae may be difficult to sample adequately by biopsy, and hard cement or necrotic bone may limit the amount of viable infected tissue obtained. Previous studies have reported culture positivity rates of infected vertebral tissue of approximately 37.5% and biopsy positivity rates of approximately 66.7% [26, 27]. For tuberculous spondylitis, biopsy site and sampling method influence diagnostic yield, with paraspinal tissue and surgical cultures often yielding higher positivity than vertebral needle biopsy [28]. Blood culture may also be useful in pyogenic spondylitis, and concordance between blood and tissue cultures has been reported to be high when both are positive [29]. Metagenomic next-generation sequencing may further improve etiological diagnosis in culture-negative spinal infection [30].

Antibiotic treatment in this cohort was therefore guided by culture results when available, pathological findings, imaging features, inflammatory markers, and clinical response. Patients with spinal tuberculosis received standard quadruple anti-tuberculosis therapy with isoniazid, rifampicin, pyrazinamide, and ethambutol during hospitalization, followed by oral anti-tuberculosis therapy for 12–18 months after discharge. For non-specific spinal infection, broad-spectrum regimens such as cefoperazone-sulbactam, piperacillin-tazobactam, carbapenems, moxifloxacin, amikacin, vancomycin-containing regimens, and linezolid-containing regimens were used according to clinical circumstances. Because many cases were culture-negative, these regimens reflect real-world management rather than a standardized protocol.

The need for broad-spectrum agents such as vancomycin or linezolid should be considered carefully. In the absence of confirmed resistant organisms, empirical broad coverage may be justified only in selected severe cases, after previous antibiotic exposure, or when Gram-positive implant-associated infection is strongly suspected. In daily practice, we attempted to narrow antibiotics when culture or susceptibility results were available. However, because this was a retrospective study, consultation patterns, treatment duration, drug monitoring, and adverse-event recording were not uniform. These issues should be addressed prospectively in future work.

Treatment of bone cement in the vertebral body responsible for infection

The treatment of vertebral infection after PKP should balance infection control, spinal stability, neurological safety, and physiological tolerance. Conservative antibiotic treatment may relieve symptoms in selected patients after sufficient anti-infective therapy [31–37]. However, many patients in this clinical scenario are elderly and experience severe persistent back pain, instability, or progressive vertebral destruction. Prolonged bed rest may also increase the risk of pulmonary infection, deep vein thrombosis, pressure ulcer, and functional decline. Previous studies have therefore supported early and active revision surgery in appropriately selected patients [38].

In the current cohort, surgical treatment was associated with a lower final ODI score than conservative treatment. Although clinically relevant, this observation does not establish a causal advantage of surgery.The choice between surgery and conservative treatment was influenced by physiological reserve, comorbidities, neurological status, spinal stability, infection severity, and patient preference. Although baseline variables were compared, the small sample size and retrospective design precluded robust confounder adjustment or propensity-score matching. The ODI comparison was therefore mainly descriptive.

All surgical procedures in this study used a posterior approach. This strategy was chosen because many patients were elderly or medically fragile, and posterior stabilization can restore mechanical stability while avoiding the additional morbidity of anterior or combined approaches. Previous studies have suggested that anterior debridement may not always be necessary for thoracolumbar tuberculosis or pyogenic spondylodiscitis, and that posterior long-segment fixation combined with short-segment fusion can achieve satisfactory results in selected patients [38–40]. The likely benefits include improved segmental stability, reduced mechanical pain, correction or prevention of deformity progression, and improved conditions for infection control under systemic antibiotic therapy.

Cement retention and the two fixation-only cases

Two patients in this cohort underwent posterior fixation without extensive lesion debridement. These cases do not imply that debridement can be routinely omitted. They were highly selected patients in whom mechanical instability and persistent pain may have been the dominant clinical problems, while systemic antibiotic therapy had already contributed to infection control. Posterior fixation may have improved spinal stability, reduced micromotion at the infected segment, and indirectly supported infection control by improving local mechanical conditions. Because only two such cases were observed, this experience remains anecdotal. It does not contradict the general principle that debridement is important when abscess, necrotic tissue, epidural compression, or progressive bone destruction is present.

Limitations

Several limitations should be considered when interpreting these findings. First, this was a retrospective study spanning 10 years, during which diagnostic work-up, antibiotic selection, surgical decision-making, and follow-up assessment were not completely uniform. Second, this study was not a pure incidence analysis. Only 3 infections occurred after PKP performed at our institution, whereas 36 patients were referred from other hospitals. Therefore, the 39-patient cohort should be regarded as a referral-enriched management cohort rather than a denominator-defined institutional cohort. Third, complete pre-PKP data were available in only about half of the patients, which limited our ability to determine whether infection was truly postoperative or had been missed before kyphoplasty. Fourth, microbiological confirmation was achieved in only a minority of patients, raising the possibility of pathogen misclassification, especially among culture-negative non-specific infections. Fifth, treatment was not randomly assigned, and the small sample size did not allow propensity-score matching or multivariable adjustment. Accordingly, the observed ODI difference cannot be interpreted as causal evidence of treatment superiority. Sixth, the cement-retention subgroup included only 8 patients and lacked a matched control group; therefore, cement retention should be viewed as a possible option in selected cases rather than as an established standard strategy. Finally, follow-up duration varied substantially, and recurrence-free survival could not be reliably assessed using time-to-event methods.

Future directions

Future studies should aim to address these limitations through prospective, multicenter designs with standardized diagnostic criteria and treatment protocols. Improved microbiological techniques, including molecular diagnostics, may enhance pathogen detection and guide antibiotic therapy. In addition, larger cohorts are needed to evaluate the safety and efficacy of cement retention strategies and to identify patient subgroups most likely to benefit from different treatment approaches.

Conclusion

Post-PKP vertebral infection is uncommon but clinically challenging and often complicated by diagnostic uncertainty. Surgical debridement and stabilization may be useful in selected patients. Partial cement retention may be feasible in carefully selected cases, but these findings require confirmation in larger prospective studies.

Authors’ contributions

Bin Zhang and Run-sheng Guo designed the experiments. Bang-lin Xie, Jiang-yuan Xiong, and Qi-hua Qi collected the clinical data. Bang-lin Xie performed data analysis. Bang-lin Xie wrote the paper. All authors have read and approved the final manuscript.

Funding

This project was supported by the Foundation of Health Commission of Jiangxi Province(20203120)and China Scholarship Council (No. 202306820035).

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee (Ethical approval No. (2022) YYL-K (2026496)) and complies with all regulations. In accordance with the ethical principles outlined in the Declaration of Helsinki, all participants provided informed consent before participating in the study. The anonymity and confidentiality of the participant were guaranteed, and participations was completely voluntary.

Consent for publication

All patients gave written informed consent for their clinical details and identifying images to be published in this study.

Competing interests

The authors declare no competing interests.