Surgical Infection after Posterolateral Lumbar Spine Arthrodesis: CT Analysis of Spinal Fusion

Abstract

Objective

To determine the incidence of infection after instrumented lumbar spine surgery, the demographic and surgical variables associated with acute infection, and the influence of infection and debridement on the consolidation of spinal fusion.

Methods

After obtaining approval from the hospital ethics committee, an observational study was made on a prospective cohort of consecutive patients surgically treated by posterolateral lumbar spine arthrodesis (n = 139, 2005–2011). In all cases, the minimum follow‐up period was 18 months. The following bivariate analysis was conducted of demographic and surgical variables: non‐infection group (n = 123); infection group (n = 16). Fusion rates were determined by multislice CT. Logistic regression analysis was performed.

Results

Incidence of deep infection requiring debridement: 11.51% (95% confidence interval, 5.85–17.18]). Bivariate analysis: differences were observed in hospital stay (7.0 days [range, 4–10] vs 14.50 days [range, 5.25–33.75]; P = 0.013), surgical time (3.15 h vs 4.09 h; P = 0.004), body mass index (25.11 kg/m² [22.58–27.0] vs 26.02 kg/m² [24.15 to 29.38]; P = 0.043), Charlson comorbidity index (median, 0 vs 1; P = 0.027), and rate of unsuccessful consolidation according to CT (18.4% vs 72.7%; P = 0.0001). In a model of multivariate logistic regression, taking as the dependent variable unsuccessful arthrodesis after 1 year, and adjusting for the other independent variables (infection, body mass index, Charlson comorbidity index, and surgical time), the only variable that was significantly associated with an outcome of unsuccessful spinal fusion after 1 year was infection, with OR = 12.44 (95% confidence interval, 2.50–61.76).

Conclusion

Deep infection after instrumented lumbar spine arthrodesis is a common complication that compromises the radiographic outcome of surgery. Patients who develop a postoperative infection and require debridement surgery are 12 times less likely to achieve satisfactory radiological fusion.

Introduction

Posterolateral lumbar arthrodesis is the procedure of choice in the treatment of certain degenerative and traumatic diseases of the lumbar spine, and is conducted to reach a solid spinal fusion, defined as the presence of trabecular bone between adjacent vertebral segments. Through spinal fusion, alignment is maintained, mechanical or segmental instability corrected, neurological deficit prevented, and pain alleviated1, 2, 3.

To achieve successful bone fusion, the diamond concept has been proposed, in which the elements of bone repair (vascularization, osteogenic cells, osteoconductive scaffold, biomechanical environment, and growth factors) are interrelated4. In applying this concept to spinal disease, it is also essential to perform an appropriate technique, enhancing fusion with bone graft and, eventually, inducing osteogenesis with bone‐forming proteins5, 6. Furthermore, osteosynthesis must be stabilized by means of specific instruments (pedicle screws and lateral bars)7.

Infection of the surgical site following spinal arthrodesis is a serious complication that occurs in 2%–15% of cases and is associated with various sociodemographic and surgical risk factors8, 9, 10, 11. The most common form of treatment for acute infection is surgical debridement, retaining the implanted instrumental arthrodesis, and obtaining intraoperative cultures to ensure appropriate antibiotic treatment and multidisciplinary management12.

The aim of this study was to determine the influence of acute postoperative infection, treated by surgical debridement, on bone fusion. As a hypothesis, we believe that during surgical debridement the diamond bone‐repair concept may be compromised due to the withdrawal of osteoinductive substances and osteoconductive tissue, and, therefore, the spinal fusion, too, would be jeopardized.

Methods

Patients

We conducted observational study of a prospective cohort of patients treated in our department between January 2005 and November 2011 by instrumented posterolateral lumbar spine fusion.

Informed consent was obtained from all individual participants included in the study. The human subjects Institutional Review Board of the Institution approved this study. All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1964, as revised in 2013.

Inclusion and Exclusion Criteria

Inclusion criteria: (i) patients had degenerative or trauma lumbar spine condition; (ii) had undergone instrumented posterolateral spine arthrodesis; and (iii) with complete preoperative data and at least 18 months follow‐up.

Exclusion criteria: (i) patients who were not followed up at our hospital; (ii) patients who had revision surgery or those for whom the surgery was the treatment for a previous infection (spondylodiscitis or paravertebral abscess) or neoplasia; and (iii) acute spinal cord injury patients (ASIA scale C or higher) treated perioperatively due to trauma or surgery.

Variables

Demographic Variables

For every patient, the demographic variables examined included age, gender, surgical risk according to the American Society of Anesthesiology scale, body mass index (BMI), preoperative diagnosis (lumbar canal stenosis, disc disease, spondylolisthesis, or fracture), smoking, and comorbidity diagnosis according to Charlson et al. modified by Deyo et al.13.

Surgical Variables

The surgical variables examined included type of fusion (including number of arthrodesis levels), surgical time (hours, minutes), and hospital stay (days). The variables related to infection were diagnosis delay, the number of surgical debridements performed, and diverse analytical variables.

Outcome Variable

The outcome variable was acute deep infection that required surgical debridement, defined by the Center for Disease Control and Prevention (Table 1) as an infection onset within the first 90 days after the surgical procedure which affects the deep soft tissue of the incision (e.g. fascial and muscle layer)14. In addition, at least one of the following criteria must be met: (i) purulent drainage from the deep incision; (ii) deep incision that spontaneously dehisces or is deliberately opened by the surgeon when the patient presents at least one of fever (>38°C), localized pain, or tenderness unless cultures are negative; and (iii) an abscess or other evidence of infection affecting the deep incision, apparent on physical examination, during reoperation, or by histopathologic or radiologic examination.

Table 1.

Diagnostic criteria for infection, according to CDC. January 201414

Infection occurs within 90 days after the operative procedure and involves deep soft tissues of the incision (e.g. fascial and muscle layers) and patient has at least one of the following: Purulent drainage from the deep incision A deep incision that spontaneously dehisces or is deliberately opened by a surgeon, attending physician and is culture‐positive or not cultured and patient has at least one of the following signs or symptoms: fever (>38°C); localized pain or tenderness. A culture‐negative finding does not meet this criterion. An abscess or other evidence of infection involving the deep incision that is detected on direct examination, during invasive procedure or by histopathologic examination or imaging test.

CDC, Centers for Disease Control and Prevention.

Fusion Evaluation

Evidence of radiological fusion was determined following Lenke's criteria for posterolateral fusion (Table 2), adapted to include multislice CT informed by a single radiologist, with each patient being classified by the quality of spinal fusion according to one of the four grades proposed by Lenke15.

Table 2.

Radiographic criteria for posterolateral fusion according to Lenke15

Grade A: Definitely solid, with solid big trabeculated bilateral fusion masses. Grade B: Possibly solid, with unilateral large fusion mass and contralateral small fusion mass. Grade C: Probably not solid, with small, thin fusion masses bilaterally. Grade D: Definitely not solid, with graft resorption bilaterally or fusion mass with obvious bilateral pseudarthrosis.

To facilitate statistical analysis and clinical interpretation, the patients were classified into two groups: grade A + B, acceptable radiological fusion; or grade C + D, unacceptable radiological fusion. When there was doubt about which of the two grades to classify a patient in, the patient was included in the lower group (C + D). We conducted a substudy as we estimate that simplifying the classification of posterolateral fusion will lead to a higher inter‐observer agreement. CT scans of 30 randomly chosen patients were assessed by four observers (three orthopaedic consultants and one radiologist). CT scans were classified according to Lenke's criteria for posterolateral fusion. Inter‐observer reliability of the four‐grade Lenke classification (A to D) and the simplified classification of posterolateral fusion into “acceptable radiological fusion” and “unacceptable radiological fusion” were calculated. We interpreted the kappa value coefficient according to the guidelines proposed by Landis and Kocch: less than 0.00, poor reliability; 0.00 to 0.20, slight reliability; 0.21 to 0.40, fair reliability; 0.41 to 0.60, moderate reliability; 0.61 to 0.80, substantial agreement; and 0.81 to 1.00, almost perfect agreement16.

Surgical Technique

Primary surgery. Surgery was performed by the same surgical team and using the same instrumentation (Xia Spinal System, Stryker Spine, Allendale, NJ, USA). The patient usually entered hospital on the day scheduled for the intervention, or on the day trauma occurred, and in the latter case the intervention was carried out within 72 h to improve the results of the reduction and to minimize complications17. Surgery was performed with the patient in the prone position, freeing the abdomen from compression and protecting the bony prominences. Antibiotic prophylaxis with Cefazolin 2 g was used as a single dose during anesthetic induction, and always at least 30 min before skin incision.

After skin preparation with 70% alcohol solution and preparation of the surgical field with iodine solution (Betadine), a posterior midline to the spine was carried out, exposing the spinal levels to be fused. Polyaxial pedicle screws were inserted at the appropriate levels, using an image intensifier to identify the entry points, performing any neurosurgical procedure when necessary, and finally preparing the host bone for grafting (facet joints, laminae, and transverse processes). The fusion was completed using bone bank allograft (variable quantity depending on the number of levels of arhtrodesis) and thawed with saline containing gentamicin 240 mg. This technique does not increase infection rates after spinal surgery18. Bone morphogenetic protein‐7 (BMP‐7; Osygraft, Stryker, Massachutetts, MA, USA) was also added in all cases6, 19. Each unit of BMP‐7 contained 3.5 mg of the rhBMP‐7 mixed with 1 g of type I bovine‐derived collagen. The total volume per unit was approximately 4 mL. If necessary, due to the extensive instrumentation employed, DTT connectors were used to make the assembly more stable. Closure was performed, leaving a 1/14 inch Redon drainage to a suction bottle, which was removed after 24 h.

Postoperative Management

Patients were allowed to sit up and to walk with assistance when their general condition permitted (usually at 48 h post‐surgery). They were discharged from hospital when the surgical wound looked good, postoperative pain was controlled with oral analgesia, and the patient could perform basic functions independently. In all cases, an elastic lumbar support was worn during the first 2 months after surgery when sitting or walking. No postoperative antibiotics were used, but for at least 3 weeks the patient was given weight‐adjusted doses of low‐molecular‐weight heparin as prophylaxis against deep vein thrombosis.

Debridement Surgery

Once the clinical diagnosis of infection was made, the patients were reoperated on according to the following protocol: skin disinfection and surgical field were prepared in the same way as for primary surgery. Meticulous resection of all necrotic tissue and loose bone allograft was carried out. Whereas well‐compacted allograft remained in place, during the debridement surgery bone morphogenetic protein 7 could not be distinguished. Primary closure was possible to accomplish in all cases and suction Redon drainage was left in place for at most 48 h. All patients needed daily regular changes of the wound dressing.

Statistical Analysis

Statistical analysis was performed using SPSS v.20 software (IBM, Armonk, NY, USA), setting the level of statistical significance at P < 0.05. A descriptive analysis was performed to determine position measurements (average and interquartile range) and the frequency distribution for qualitative variables. The cumulative incidence of infection was calculated for a 95% confidence interval (95% CI). Taking the presence of infection as the comparison variable, the χ²‐test was used for the qualitative variables (the linear association test for ordinal qualitative variables) and the Mann–Whitney U‐test for the quantitative independent variables. Finally, a multivariate logistic regression model was created, taking incomplete fusion at 1 year as the dependent variable, and adjusting for independent variables such as infection, BMI, Charlson index, and surgical time, calculating the odds ratios with 95% CI.

Results

Descriptive Analysis

Of the 156 patients selected from our surgical registry of spinal arthrodesis, 17 were excluded from the study: 4 patients who presented a positive culture in the surface drainage from the surgical wound, but whose subsequent treatment with antibiotics and wound care was satisfactory, with no deep infection being observed; 7 patients for whom posterolateral fusion surgery was performed as treatment for a prior infection (spondylodiscitis); 2 patients who received revision surgery of the arthrodesis; 2 patients who received percutaneous surgery with no attempt to achieve spinal fusion; 1 patient who died in the immediate postoperative period due to severe acute respiratory failure; and 1 who suffered perioperative paraplegia that required transfer to the spinal cord injury center.

Demographic and Surgical Data

Overall, 139 patients met the inclusion criteria. The demographic and surgical data of these patients are summarized in Table 3. The average age was 43 years (interquartile range [IQR], 35–55), with 74 men (53.23%) and 65 women (46.76%). The average BMI was 25.53 kg/m² (IQR, 23.14–27.54). The average comorbidity score, according to the Charlson index, was 0.48 (SD, 0.805). The surgical risk according to the American Society of Anesthesiology scale was I in 58 cases (42.6%), II in 71 (52.2%) and III in 7 (5.1%), while in 3 cases this parameter was not reflected in the medical record. Among the patients, 29.5% were smokers and 8.6% were diabetics being treated with oral antidiabetic agents or insulin.

Table 3.

Descriptive data for entire cohort (n = 139)

Variable	Data
Age (years)	43 (IQR: 35–55)
Sex
Male	74 (53.2%)
Female	65 (46.8%)
BMI (kg/m2)	25.53 (IQR: 23.14–27.54).
Smoker	41 (29.5%)
Diabetes	12 (8.6%)
Charlson	0.48 (SD 0.81)
ASA risk
I	58 (42.6%)
II	71 (52.2%)
III	7 (5.1%)
N/A	3 (2.2%)
Diagnosis
Fracture	53 (38.1%)
Disc disease	53 (38.1%)
Lumbar canal stenosis	23 (16.5%)
Spondylolisthesis	10 (7.2%)
Surgery time (min)	204 (IQR: 165–256)
Year of surgical intervention
2005	22 (15.8%)
2006	24 (17.3%)
2007	16 (11.5%)
2008	8 (5.8%)
2009	20 (14.3%)
2010	31 (22.3%)
2011	18 (12.9%)
Fusion levels	Median: 2
1	26 (18.7%)
2	60 (35.5%)
3	16 (11.51%)
4	26 (18.7%)
5	8 (5.7%)
6	2 (1.43%)
7	1 (0.71%)
Annual CT	98 (70.1%)
Infection	16 (11.51%)
Consolidation
Successful (A–B)	74 (75.5%)
Unsuccessful (C–D)	24 (24.5%)
Lost to follow up	2 (1.4%)
Surgical material removed	2 (1.4%)

ASA, American Society of Anesthesiologists; BMI, body mass index.

The preoperative diagnosis justifying spinal surgery referred to a fracture in 53 patients (38.1%), disc disease in 53 (38.1%), stenosis in 23 (16.5%) and spondylolisthesis in 10 (7.2%). The median number of levels fused was two (range, 1–7) and the average duration of surgery, including patient positioning, was 204 min (IQR, 165–256 min).

Evaluation Results of Fusion

The annual CT test (range, 12–18 months) to assess the fusion was performed in 98 patients (70.1%) and spinal fusion was considered satisfactory (group A or B, according to Lenke) in 74 patients (75.5%) and unsuccessful in the remaining 24 patients (24.5%)15.

The inter‐observer and intra‐observer agreements of the radiological fusion classified by all observers using the four‐grade Lenke classification and simplified classification for both trauma surgeons and radiologist, are presented in Tables 4 and 5.

Table 4.

Intra‐observer agreement

Observer	Original Lenke classification			Modified Lenke classification
	Kappa	P‐value	%	Kappa	P‐value	%
1	0.338	0.003	53.3	0.595	0.001	83.3
2	0.529	0.001	66.7	0.593	0.001	80.0
3	0.691	0.001	83.3	0.864	0.001	93.3
4	0.639	0.001	76.7	0,792	0,001	93,3
Mean	0.549	0.001	70	0.711	0.001	87.5

Table 5.

Interobserver agreement

Observer	Original Lenke classification			Modified Lenke classification
	Kappa	P‐value	%	Kappa	P‐value	%
1 vs 2	0.326	<0.001	50.0	0.298	0.068	63.3
1 vs 3	0.118	0.273	36.7	0.282	0.087	66.7
1 vs 4	0.277	0.017	50.0	0.706	0.001	90.0
2 vs 3	0.092	0.408	36.7	0.345	0.050	66.7
2 vs 4	0.200	0.041	40.0	0.321	0.017	63.3
3 vs 4	0.192	0.082	43.3	0.203	0.197	63.3
Mean	0.201	0.137	42.8	0.359	0.07	68.9

Diagnosis and Management of Infection

Of the 139 patients analyzed, 16 patients were diagnosed with postoperative deep infection and required debridement surgery (Fig. 1). The cumulative incidence of infection in the cohort was 11.51% (95% CI, 5.85%–17.18%). For the analysis, the patients were divided into the following groups: non‐infection group, n = 123; and infection group, n = 16. Table 6 shows the demographic, surgical, and infection characteristics of the patients in the infection group. No annual or quarterly development of infection was observed.

Figure 1.

Clinical photograph shows purulent accumulation over instrumentation in patient with posterolateral arthrodesis.

Table 6.

Descriptive data for infection group (n = 16)

No.	Gender	Age(years)	BMI (kg/m2)	Charlson	Diagnosis	Surgery time	Levels	Hospital stay (days)	Readmission	Debridements	Delay	Germ	Fusion
1	F	46	29.29	0	Fracture	3:00	3	37	No	1	7	E. cloacae	—
2	F	59	38.93	1	Stenosis	4:20	3	278	No	5	10	Polymicrobial	IR
3	M	34	24.34	0	Discopathy	3:45	1	63	No	2	15	Polymicrobial	IR
4	F	50	22.26	2	Stenosis	3:10	5	63	No	1	7	Polymicrobial	Yes
5	F	49	25.79	0	Spondylolisthesis	3:00	1	53	No	1	7	S. aureus	No
6	F	66	23.13	1	Spondylolisthesis	3:25	2	40	Yes	3	11	E. coli	—
7	F	34	27.68	0	Stenosis	4:00	1	38	Yes	1	30	Polymicrobial	Yes
8	M	49	26.26	2	Discopathy	4:40	2	11	Yes	1	17	—	No
9	F	41	29.76	0	Stenosis	4:45	2	25	Yes	1	10	S. aureus MR	No
10	M	39	29.41	0	Stenosis	4:25	2	25	No	1	13	Polymicrobial	—
11	F	29	32.32	0	Discopathy	5:05	2	24	Yes	1	12	S. agalactiae	No
12	M	77	24.09	2	Stenosis	4:25	4	62	No	1	14	E. coli	—
13	M	53	23.12	2	Fracture	3:45	4	18	Yes	1	14	Polymicrobial	—
14	F	55	29.00	1	Discopathy	4:30	2	46	No	1	12	P. mirabilis	—
15	M	40	25.43	1	Fracture	3:40	4	33	Yes	1	23	S. hyicus	No
16	F	66	25.40	1	Stenosis	5:30	4	34	Yes	1	17	E. coli	Yes

BMI, body mass index; F, female; IR, implant removal; M, male; —, not available.

Among the group of patients who developed an infection, the mean time to its diagnosis was 12.5 days (range, 7–30) so all cases were diagnosed as acute infection. The average length of hospital stay after debridement surgery was 23 days (15–28.75). In 15 patients, the agent responsible for the infection was cultured, and found to be gram negative in 31.25% of the cases (E coli, 3 cases; E. cloacae, 1 case; P mirabilis, 1 case), followed by methicillin‐sensitive Staphylococcus aureus (MSSA) in 1 patient and methicillin‐resistant Staphylococcus aureus (MRSA) in another patient. Other gram positive strains that were isolated were S agalactiae in one case and S hyicus in another. In one patient, the strain responsible could not be identified (negative culture). In six patients (37.5%) the culture was polymicrobial, containing a gram‐positive cocci in every case, and, thus, the gram‐positive infections accounted for 62.5% of all infections. All patients received intravenous antibiotic treatment after surgery while they remained hospitalized and on discharge; they were prescribed oral antibiotic therapy in accordance with the antibiogram results. All contact with these patients was conducted in conjunction with the hospital's infectious disease unit, and the analytic parameters used corresponded to standard acute‐phase reactants.

Of these 16 patients, 13 (81.25%) required only one surgical debridement to control the infection, but another 3 needed 2, 3 and 5 such debridements, respectively. In 1 of these cases, in which 5 debridements were needed, it was necessary to perform adjuvant treatment consisting of negative‐pressure wound therapy, removal of instrumentation and case management by the plastic surgery department due to skin coverage problems; despite these measures, infection control was uncertain until present (8 years after surgery)20; there is no evidence of fistula and acute phase reactants are negative, but the functional outcome of the patient is poor.

Bivariate Analysis

The bivariate analysis between the two groups produced the following results (see Table 7): for the non‐infection group (n = 123) vs the infection group (n = 16) there were statistically significant differences in hospital stay (7.0 days [range, 4–10] vs 14.50 days [range, 5.25–33.75]; P = 0.013), surgical time (3.15 h vs 4.09 h; P < 0.01), BMI (25.11 kg/m² [22.58–27.0] vs 26.02 kg/m² [24.15–29.38]; P = 0.043), Charlson comorbidity index (median, 0 vs 1; P = 0.027), and rate of unsuccessful consolidation (18.4% vs 72.7%; P < 0.01).

Table 7.

Inter‐group comparison (n = 139)

	Non‐infection group	Infection group	P‐value
	(n = 123)	(n = 16)
Age (years)	43.0 (34.75–55)	49.45 (39.41–58.04)	0.223
Sex
Male	55.38%	37.5%	0.280
Female	44.62%	62.5%
Hospital stay (d)	14.50 (4–10)	7.0 (5.25–33.75)	0.013
Fusion levels	2 (2–4)	2 (2–4)	0.771
Surgery time (h)	3:15	4:09	0.004
ASA risk			0.361
I	44.2%	31.3%
II	50.8%	62.5%
III	5.0%	6.3%
IV	0%	0%
BMI (kg/m2)	25.11 (22.58–27.0)	26.02 (24.15–29.38)	0.043
Smoker	29.16%	31.25%	0.864
Diabetes	7.5%	12.5%	0.619
Charlson	0	1	0.027
Unsuccessful consolidation	18.4%	72.7%	<0.001

ASA, American Society of Anesthesiologists; BMI, body mass index.Bold values are statistically significant, p<0.05.

In our analysis of the risk of unsuccessful spinal fusion after infection, compared to uninfected patients the patients in the infection group had a relative risk of 3.95 (95% CI, 2.23–7.01) and an OR of 11.83 (95% CI, 2.82–49.62) of not achieving a successful fusion.

Multivariate Analysis

To avoid the influence of confounding variables, a multivariate logistic regression model was created, taking as the dependent variable unsuccessful arthrodesis at 1 year, and adjusting for the other independent variables (infection, BMI, Charlson index, and surgical time). In this model, the only variable that was significantly related to unsuccessful spinal fusion at 1 year was the infection variable, with an OR of 12.44 (95% CI, 2.50–61.76).

Discussion

This paper analyzes a cohort of consecutive patients (n = 139) who each underwent posterolateral arthrodesis of the lumbar spine, as treatment for lumbar spinal disease, during a specific time period (2005–2011). We compared patients who required surgical debridement (n = 16) to those who remained free from infection (n = 123) and, therefore, did not require surgery that would have influenced the initial objective of carrying out a spinal fusion procedure. We studied bone consolidation, measured by CT, at 1 year after surgery and analyzed variables related to postoperative deep infection.

In the present study, the incidence of deep infection after posterolateral spinal arthrodesis was 11.51%. Although this figure is in line with the findings of other studies (5%–15%)9, 10, 11, we consider this value rather high for surgery that in approximately 50% of our cases is performed electively. The wide range of incidence reported in previous studies is probably due to the inclusion in their analyses of other types of spinal surgery with a lower risk of infection (e.g. cervical surgery, spine surgery without instrumentation or kyphoplasty)10.

In the present study, the following variables were found to significantly influence the development of infection: operating time (major surgery lasting over 3 h), comorbidity with the presence of at least one Charlson comorbidity, and BMI > 25 kg/m² 21, 22, 23. All these variables have been the subject of previous study24, 25; other variables have also been studied, such as diabetes or smoking, but we did not find them significant, probably due to the power of our study. We did not record any monthly or annual differences suggesting a learning curve or the existence of seasonal trends26, 27.

Although in the cases analyzed, various standard prophylactic procedures were conducted to minimize postoperative infection, we believe more emphasis should be placed on measures aimed at reducing the incidence of infection, some of which are described in the current literature, and on controlling preoperative risk factors, by means such as intraoperative irrigation with abundant physiological serum or Betadine, avoiding surgical shaving, or using gentamicin microspheres28, 29, 30, 31, 32, 33, 34. Recently, moreover, the use of postoperative prophylactic antibiotics has been suggested (evidence level: C)35, 36.

Currently, the most widely accepted treatment for infection after spinal arthrodesis is surgical debridement of the wound, removal of the allograft, and the retention of instrumentation, to avoid compromising the stabilization, which is necessary to treat the infection12. It is also essential in medical management to apply intensive antibiotic therapy adjusted according to intraoperative cultures37. Nevertheless, some authors, such as Hong et al., have proposed that deep infection should be managed using only antibiotics, while Rohmiller et al. performed suction irrigation without surgery and Kim et al. removed all the material that had been implanted38, 39, 40. All of these authors reported good results. In our experience, infection should be addressed aggressively and the instrumentation retained, because in most cases (93.75% in our series) this is what enables the infection to be overcome.

Dipaola et al. recently proposed a model to predict the number of surgical debridements required, and reported that MRSA infection and bacteremia were the strongest predictors of the need for multiple surgical interventions41. In our study, 3 patients required more than one such intervention. Multiple interventions probably influence the fact that the patient does not remain free of infection, but further studies would be needed to confirm this hypothesis, with larger numbers of study subjects, and carried out over an extended period. We note, moreover, that infection resulting from spinal surgery consumes significant amounts of resources; in our series, the average length of hospital stay increased significantly, from 7 to 14.5 days.

It has been suggested that the consolidation of spinal fusion may be used as a measure of its success, but problems arise in quantifying this variable. While CT seems to be the best tool for evaluating spinal fusion, we have had some difficulty in assessing the quality and quantity of arthrodesis in small interfacetal or interspinous joints, with poor interobserver reliability and the need to assess all CT scans, especially sagittal and coronal cuts42, 43. In our experience, in the presence of several fused levels, one of which is unfused (e.g. L₃–S₁ arthrodesis with L₅–S₁ pseudarthrosis) (Fig. 2), radiological assessment was difficult and, most importantly, there was insufficient evidence of the clinical results.44 To counteract this potential bias, in our study each case was assessed by the same expert radiologist and an intra‐observer and inter‐observer substudy was conducted.

Figure 2.

Difficult assessment of nonunion by CT: L₅–S₁ nonunion with large bilateral L₂–L₅ fusion bars. (A) Coronal view. (B) Sagittal view (note: CT views of the same patient).

Poor reliability of the four‐category Lenke classification was confirmed. Better reliability was found for the classification of radiological fusion simply as “acceptable radiological fusion” and “unacceptable radiological fusion”.

Although our initial bivariate analysis revealed no significant association between the infection and non‐infection groups, a subsequent multivariate logistic regression model showed that only infection addressed by surgical debridement appeared to be related to unsuccessful fusion. No such relation was observed for the other variables (surgical time, Charlson comorbidity index, or BMI) or between them.

We conclude, therefore, that it is the infection itself or the subsequent treatment (surgical debridement, removal of the allograft and of the bone morphogenetic protein, and/or prolonged antibiotic therapy) or both factors, which may influence the consolidation, producing a significant difference in the quality of the spinal fusion (72.7% vs 18.4%).

To the best of our knowledge, this is the first paper to analyze the consolidation achieved following posterolateral arthrodesis and subsequent infection, using CT to assess the success of the spinal fusion. Weiss et al. studied a series of 29 patients who developed infection, which was treated in a way similar to that described in this paper, using dynamic plain radiographs (XR). They recorded a nonunion (pseudarthrosis) rate of 37.9%, in contrast to our own finding of 72.7% unsuccessful consolidation45. We believe this large difference is probably related to the use of CT to evaluate the consolidation, as several studies have reported that the analysis of spinal fusion bars, comparing the fusion outcomes as revealed by XR, CT, and open surgery, shows that plain radiography tends to overestimate bone consolidation due to the presence of nonunion at various levels which is not assessed correctly42, 43, 46.

This study presents certain limitations. Most importantly, due to the review of the cohorts we did not obtain sufficient information to perform a validated analysis of the clinical outcome achieved by the patients (Oswestry or SF36). Very little has been published previously regarding the long‐term evaluation of infection arising from posterolateral spinal arthrodesis. Petilon et al. (2012) evaluated a series of 30 patients with postoperative infection and found that after 2 years the patients with infection suffered more back pain than did non‐infected patients47. It would be interesting to conduct a medium‐term functional analysis of patients who develop postoperative infection.

Another potential limitation of the present study is its considerable length (7 years), otherwise necessary to obtain a sufficient sample size. In spite of this, the members of the surgical team, who performed the interventions, have followed the same surgical and infection management protocol throughout the study period. Furthermore, due to the characteristics of the study, it was not possible to apply a more evidence‐based methodology (e.g. by means of an experimental approach, with a control group), which obliges us to be very cautious in drawing conclusions.

Among the strengths of our study is the large data series analyzed (a consecutive cohort with a minimum follow‐up of 18 months), the homogeneity of the patients, the consistency of the information obtained (few losses to the study), and the CT scan performed in most cases (70.1%). Thus, despite the limitations referred to above, we can draw useful conclusions to improve the perioperative management of patients who undergo surgery for posterolateral spinal arthrodesis (personalized risk information, control of weight, and comorbidities), and to ensure appropriate clinical and radiological monitoring of patients who develop an infection.

Our data series shows that when deep infection occurs after instrumented lumbar spine surgery, one or more surgical debridements are needed to treat it; although in most cases the patient then remains free of infection, there is an increased risk of unsuccessful consolidation, which leads us to believe that in the long term this situation may influence clinical and functional outcomes (producing a rupture of the implant, residual back pain, and infected pseudarthrosis). However, further studies and longer follow‐up periods are needed to confirm this hypothesis.

In conclusion, in our study group, various demographic and surgical variables were shown to be related to postoperative infection (BMI >25 kg/m², surgical time > 3h, and Charlson comorbidity index equal or higher than 1), increasing its probability by 11.1%. The relative risk of unsuccessful radiological consolidation is 4 times greater, and the odds ratio is 12 times greater when the patient has suffered an acute postoperative infection of the posterolateral lumbar spine arthrodesis. However, the majority of patients remain free from infection when the proposed treatment (surgical debridement, removal of the implant, and subsequent antibiotic therapy) is applied.