Inter- and intraobserver variability in the postoperative evaluation of transpedicular stabilization: computed tomography versus magnetic resonance imaging

The Spine Journal 10 (2010) 285–290

Clinical Study

Inter- and intraobserver variability in the postoperative evaluation of transpedicular stabilization: computed tomography versus magnetic resonance imaging Stephan W. Tohtz, MDa,*, Patrick Rogalla, MDb, Matthias Taupitz, MDb, Carsten Perka, MDa, Tobias Winkler, MD, PhDa, Michael Putzier, MDa a

Department of Orthopedics, Center for Musculoskeletal Surgery, Charite´–Universitaetsmedizin, 10117 Berlin, Germany b Department of Radiology, Charite´–Universitaetsmedizin, 10117 Berlin, Germany Received 1 April 2009; revised 18 November 2009; accepted 25 December 2009

Abstract

BACKGROUND CONTEXT: Computed tomography (CT) represents the state of the art for the postoperative verification of the implant position after transpedicular stabilizations. Magnetic resonance imaging (MRI) has not challenged the CT, yet, because of susceptibility artifacts but would be favorable as a diagnostic tool for its excellent soft-tissue qualities. PURPOSE: A study that analyzed if an artifact-reduced MRI could overcome this problem and provide sufficient data for the postoperative assessment was conducted. STUDY DESIGN: The study design was a radiologic comparison of CT and MRI techniques evaluating pedicle screw placement after spinal fusion. PATIENT SAMPLE: Fifty consecutive patients were given an MRI and a CT after a transpedicular stabilization surgery. Thirty-eight patients suffered from degenerative spinal disorders; three surgeries had become necessary because of spondylodiscitis, eight patients suffered from metastatic vertebrae destruction, and one patient experienced a fracture. OUTCOME MEASURES: Any contact of a malpositioned pedicle screw with the dura and/or radicular structures was identified as an implant-associated complication and was compared with postoperative clinical patient findings. METHODS: In total, 338 pedicular screws were analyzed in regard to their intrapedicular position. The double-blind evaluation of MRI and CT data was carried out by two radiologists and two spine surgeons. Accuracy of the CT analysis was calculated based on the interobserver agreement of 100%. Magnetic resonance imaging accuracy was calculated. RESULTS: The interobserver accuracy of the CT data amounted to a median of 89.8% and in the MRI data of 86.7%. Intraobserver comparisons showed a significant difference between CT and magnetic resonance evaluations in one observer (k50.293). In all other observers, the results were concordant with kappa values from k50.328 to k50.702. There was a high degree of agreement regarding the diagnosis of malpositioned pedicle screw and corresponding clinical symptoms between both techniques. CONCLUSIONS: The presented data show that artifact-reduced MRI is equivalent to CT imaging in the postoperative evaluation of titanium spinal rod-screw systems. We therefore conclude that MRI should be considered as an alternative tool for the golden standard CT for postoperative imaging controls for its advantages in soft-tissue analysis. Ó 2010 Elsevier Inc. All rights reserved.

Keywords:

Transpedicular stabilization; Spine; CT; MRI; Screw placement

Introduction FDA device/drug status: not applicable. Author disclosures: none. * Corresponding author. Department of Orthopedics, Center for Musculoskeletal Surgery, Charite´–Universitaetsmedizin, Charite´platz 1, 10117 Berlin, Germany. Tel.: (þ49) 30-450-515109; fax: (þ49) 30-450515922. E-mail address: [email protected] (S.W. Tohtz) 1529-9430/10/$ – see front matter Ó 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.spinee.2009.12.020

Computed tomography (CT) represents the state of the art among imaging techniques for controlling the position of the implants after transpedicular stabilization [1–3]. By modification of parameters such as slice thickness and gantry angle and the application of filters, CT enables almost

286

S.W. Tohtz et al. / The Spine Journal 10 (2010) 285–290

Context Computed tomography (CT) scanning is currently the most commonly used modality to check screw placement in the thoracolumbar spine. This study aims to assess magnetic resonance imaging (MRI) for this indication. Contribution Interobserver CT accuracy was found to be 89.8% versus 86.7% for MRI; and high agreement regarding malpositioned screws was noted. Implication The findings suggest artifact-reduced MRI might be an alternative to CT for the assessment of screw positioning. The ‘outlyer’ (poor intra-observer correlation) suggests the technique might well be highly reader-dependent, raising some concern for the ability to generalize. —The Editors

artifact-free and high-resolution images of bone and implant outlines, so that an outstanding assessment of spinal and pedicular structures is possible as well as the evaluation of the position of the screws after transpedicular stabilization [4,5]. The drawbacks of this technique in postoperative situations are a restricted assessment of neural structures, the necessity of a secondary reconstruction of sagittal planes, and the radiation exposure [6]. Magnetic resonance imaging (MRI), as the standard preoperative imaging procedure, is indispensable for the evaluation of spinal and radicular alterations. No other imaging procedure enables statements with such a high sensitivity and specificity in the diagnosis of degenerative spine disease [7]. Also in the immediate postoperative period after spinal surgery, magnetic resonance (MR) scans provide excellent information about soft-tissue structures [8,9]. However, until now, MR-based imaging techniques are not commonly used for postoperative examinations after stabilization surgery because of the artifacts caused by metallic implants [10,11]. Consecutively, two imaging procedures with different quality features are being applied for the pre- and postoperative diagnostic evaluations in spine surgery. It has been disputed in the recent past to what extent an artifact-reduced MRI enables a statement equivalent to CTwhen using titanium implants, but a performance analysis does not exist so far [12]. For this reason, a clinical trial was conducted, which evaluated the validity of postoperative MR versus CT imaging in the analysis of pedicle screw placement.

Thirty-eight patients underwent spinal fusion because of degenerative spine disease, three patients received fusion surgery because of spondylodiscitis, and a posterior stabilization was performed in eight patients because of metastatic vertebral body destruction and in one case because of a fracture. Surgery included mono-, bi-, and multisegmental spinal fusions and was performed with a titanium rod-screw system (Allospine; Sulzer Medica, Winterthur, Switzerland). In total, 338 pedicle screws were analyzed in regard to their intrapedicular position by using a SR 7000 CT scanner (Philips, Eindhoven, The Netherlands) and a Magnetom Vision II MR scanner (Siemens, Erlangen, Germany) with maximal gradient amplitude of 25 mT/m. Helical CT scans were made with a slice thickness of 2 mm, 120 kV, and 100 mA, with the gantry angle parallel to the screws aligned through the disc space or the fused vertebral body. Both bone and soft-tissue windows were analyzed. For MR examinations, the following parameters were used: sagittal T1-weighted (repetition time 520–700 ms, echo time 10–12 ms), combined proton density, and T2-weighted (repetition time 3000–4238 ms, echo time 20 and 120 ms) turbo spin echo sequences with 350350 mm field of view; and axial T1-weighted (repetition time 900 ms, echo time 10 ms) turbo spin echo sequences with 290290 mm field of view. The slice thickness for all MR examinations was 4 mm. Minimization of susceptibility-induced artifacts by the implants was achieved by choosing settings for sampling bandwidth, inter-echo spacing, and field of view, which led to maximal gradient amplitude for signal readout. The double-blind evaluation of MRI and CT records was carried out by two radiologists (Observers 1 and 2) and two spine surgeons (Observers 3 and 4). The position of a screw within the pedicle was either judged central or misplaced. Central positioning was defined for pedicle screw deviations up to 2 mm, because deviations of this amount are generally not associated with neurological complications [3,13]. Table 1 shows the exact numbers of nonperforated, less than 2 mm perforated, and greater than or equal to 2 mm perforated screws. Deviations of more than 2 mm of the central position were documented as malposition [14,15]. Figure 1 shows axial CT and MR scans of the vertebras of two patients with well-positioned screws, nerve root compression by an S1 screw, and hematoma formation anterior to a vertebra because of anterior cortical perforation. For the statistical evaluation of the inter- and intraobserver results, the McNemar test was used. The Cohen kappa coefficient was used to judge the correlation of the results. Data are given as median and range, if not stated otherwise. Inter- and intraobserver reliability were also analyzed separately for patients stabilized for degenerative disease and nondegenerative disease (Table 2).

Materials and methods Fifty consecutive patients (27 women and 23 men) received both spinal MRI and CT after transpedicular stabilization surgery between Postoperative Days 2 and 10.

Results Analyzing the CT data of our patients’ accuracy amounted to a median of 89.8% (83.4–96.6%). Accuracy

S.W. Tohtz et al. / The Spine Journal 10 (2010) 285–290

287

Table 1 Number of screws diagnosed nonperforated, less than 2 mm perforated, or greater than or equal to 2 mm perforated by the four observers in CT and MR scans Observer 1

Observer 2

Observer 3

Observer 4

Screw position

CT

MRI

CT

MRI

CT

MRI

CT

MRI

Nonperforated Cortical perforation (!2 mm) Misplaced ($2 mm)

232 34 72

234 48 56

266 35 37

231 33 74

283 39 16

281 45 12

274 38 26

279 41 18

CT, computed tomography; MRI, magnetic resonance imaging.

of the MRI data, 86.7% (81.1–97.6%), was found to be not significantly different from the CT value. Accuracies of CT and MRI were calculated based on the interobserver agreement of 100% (Fig. 2 and Fig. 3), because only the intraoperative diagnosis is accepted as a gold standard next to CT evaluation. The MRI data have been additionally compared with the gold standard CT, demonstrating an accuracy value of 87.3% (median), which did not differ from the value obtained from the comparison with the interobserver agreement. In total, 260 screws in CT scans and 251 screws in MR scans were diagnosed centrally placed by all observers. Eleven screws were diagnosed misplaced by all observers in both techniques (Fig. 2 and Fig. 3).

The interobserver comparison of the CT data resulted in a considerable agreement (Cohen kappa coefficients amounted to a median of kM50.515 [0.310–0.696]). However, differences were found when comparing the investigators pairwise. The recorded positions (central, malpositioned) showed a significant interobserver difference between all physicians when evaluating all 338 pedicle screws (p less than .005). The pairwise comparison of the MRI data showed significant differences (p less than .05) between all but one (Observers 3 and 4) pairs of observers. The intraobserver comparison of both methods showed significant differences between CT and MRI diagnoses only for Observer 2 (kappa coefficient of k250.293; p less than .005). Concordance could be found between the results of

Fig. 1. Comparison of computed tomography and magnetic resonance imaging scans in two patients. (Top Left, Top Right) Central positioning of pedicle screws L4 with hematoma on the anterior side of the vertebra surrounding the iliac vessels caused by perforation of the tip of both screws. (Bottom Left, Bottom Right) Medial misplacement of the right pedicle screw S1 with nerve root compression.

288

S.W. Tohtz et al. / The Spine Journal 10 (2010) 285–290

Table 2 Computed tomography and magnetic resonance imaging inter- and intraobserver reliabilities analyzed separately for patients with degenerative and nondegenerative diseases All patients, kM(338 screws)

Thirty-eight patients (degenerative), kM(254 screws)

Twelve patients (nondegenerative), kM(84 screws)

0.487 0.274

0.623 0.467

300 250 200 150 100

Interobserver comparison CT 0.515 MRI 0.327

50 0

Intraobserver comparison Observer 1 0.328 Observer 2 0.293 Observer 3 0.702 Observer 4 0.563

1

0.327 0.276 0.783 0.596

n 350 300 250 200 150 100 50 0 3

central central with interobserver agreement misplaced misplaced with interobserver agreement

3

4

central central with interobserver agreement misplaced misplaced with interobserver agreement

the evaluation of CT and MRI data in the other observers (Table 2). The separate analysis of the inter- and intraobserver reliabilities of the patients with degenerative and nondegenerative spinal diseases is shown in Table 2. The exclusion of patients with stabilization surgery because of tumor, spondylodiscitis, or fracture would not significantly alter the results. The documentation of associated pathological morphological findings showed clear differences between the CT and the MRI. The postoperative MRI showed a higher number of hematomas (p less than .001) with or without contact to neural structures and also a higher number of radicular alterations (p less than .05). There was no difference between the two imaging techniques in terms of the frequency of diagnosed spinal stenosing. Three patients (13, 14, and 46) were diagnosed with new, postoperatively occurring radicular symptoms (2 sensitive, 1 motor), which were caused by malpositioned screws. These were noticed in both the CT and the MR scans. In these patients, nerve root compression

2

2

observer

0.385 0.375 0.538 0.493

CT, computed tomography; MRI, magnetic resonance imaging.

1

n 350

4 observer

Fig. 2. Interobserver comparison of Observers 1 to 4 for assessment of findings in computed tomography. Differentiation in central and misplaced pedicle screws with significant differences between all observers (p less than .05). Level of 100% interobserver agreement: n5260 (central) and n511 (misplaced).

Fig. 3. Interobserver comparison of Observers 1 to 4 for assessment of findings in magnetic resonance imaging. Differentiation in central and misplaced pedicle screws with significant differences between all observers (p less than .05), except for the paired comparison of Observers 3 and 4. Level of 100% interobserver agreement: n5251 (central) and n511 (misplaced).

and hematoma formation could be detected in the MR scans more than two times more frequent than in the CT scans (Table 3). In one case (Patient 5), a preoperatively existing sensomotor deficit with postnucleotomy syndrome progressed. In this instance, the two imaging techniques did not show any implant-related causes. Nevertheless, significant local hematoma formation was detected by all observers in the MR scans of this patient. Discussion In the present study, the validity of artifact-reduced MRI in evaluating screw placement after spinal fusion surgery is analyzed in comparison with CT. Previous MRI investigations judging postoperative pedicle screw positions or screw channels after removal of the implant have so far been unable to establish values for the validity of MR scans [12,16]. Although there is concordance concerning the definition and classification of misplaced pedicle screws [4,10,17], differing statements about the prevalence of screw misplacement can be found when comparing the results of some authors [1,2]. In the present study, misplacement was diagnosed only in cases where the screw had perforated 2 mm or more from the pedicle wall. The median of misplaced screws of all patients as diagnosed on the CT images (9.3%) and MR images (10.9%) correlated with the literature of CT-evaluated data on transpedicular stabilization [3,4,18]. Interestingly, a clear interobserver variability could be detected in both the CT-based and MRI-based analysis of screw positions, whereas the intraobserver comparison of the two methods did not display differences in all but one observer. These results demonstrate that, obviously, the quality of the diagnosis of screw positioning after spinal surgery is influenced more by the observer than by the method.

S.W. Tohtz et al. / The Spine Journal 10 (2010) 285–290

289

Table 3 Number of patients with newly occurring or progressing neurological deficits and frequency of associated radiological abnormalities Patient

Surgery performed

Postoperative neurological status

Accuracy of CT findings (observer frequency)

Accuracy of MRI findings (observer frequency)

5

ALIF and transpedicular instrumentation L4–S1 Transpedicular instrumentation L1–S1

Progression of L5 radiculopathy right L4 radiculopathy right (sensible)

—

14

ALIF and transpedicular instrumentation L4–S1

S1 radiculopathy right (sensible)

Misplaced screw S1 (4)

46

Transpedicular instrumentation L2/L3–L5/S1

S1 radiculopathy right (motoric)

33

Transpedicular instrumentation T12–S1

Unspecific leg pain left

Misplaced screw L5 (4) Hematoma (1) Nerve root compression (2) Hematoma (1)

Misplaced screw L4 (1) Hematoma (4) Misplaced screw L4 (4) Hematoma (4) Nerve root compression (3) Misplaced screw S1 (4) Hematoma (2) Nerve root compression (3) Misplaced screw L5 (4) Hematoma (4) Nerve root compression (4) Hematoma (1)

13

Misplaced screw L4 (4) Nerve root compression (1)

CT, computed tomography; MRI, magnetic resonance imaging; ALIF, antero-lateral interboby fusion.

The accuracy of artifact-reduced MRI corresponded approximately to the range of accuracies reported in CT evaluations of thoracolumbar screw placement (79.0–94.3%) [1,2,19,20]. This accuracy had been accomplished with a slice thickness of the MR scans of 4 mm, which was double the value of the CT scans. As we did not observe differences in the numbers of screws diagnosed misplaced between both methods, we do not believe that the used slice thickness contributed to missed malpositions. In accordance with the existing literature in our study, the rate of neural or vascular complications after transpedicular stabilization surgery was considerably lower than the rate of misplaced screws [21–23]. Taking this into account and including the fact that MRI allows a safe assessment of neural and surrounding softtissue structures [24–26], we conclude that artifact-reduced MRI should be considered as an alternative tool for the golden standard CT for postoperative imaging controls after spinal fusion surgery.

References [1] Kosmopoulos V, Schizas C. Pedicle screw placement accuracy: a meta-analysis. Spine 2007;32:E111–20. [2] Yoo JU, Ghanayem A, Petersilge C, Lewin J. Accuracy of using computed tomography to identify pedicle screw placement in cadaveric human lumbar spine. Spine 1997;22:2668–71. [3] Belmont PJ Jr, Klemme WR, Dhawan A, Polly DW Jr. In vivo accuracy of thoracic pedicle screws. Spine 2001;26:2340–6. [4] Upendra BN, Meena D, Chowdhury B, et al. Outcome-based classification for assessment of thoracic pedicular screw placement. Spine 2008;33:384–90. [5] Tehranzadeh J, Gabriele OF. The prone position for CT of the lumbar spine. Radiology 1984;152:817–8. [6] Mahnken AH, Wildberger JE, Gehbauer G, et al. Multidetector CT of the spine in multiple myeloma: comparison with MR imaging and radiography. AJR Am J Roentgenol 2002;178:1429–36. [7]. van Rijn JC, Klemetso N, Reitsma JB, et al. Observer variation in the evaluation of lumbar herniated discs and root compression: spiral CT compared with MRI. Br J Radiol 2006;79:372–7.

[8] Mirzai H, Eminoglu M, Orguc S. Are drains useful for lumbar disc surgery? A prospective, randomized clinical study. J Spinal Disord Tech 2006;19:171–7. [9] Morse K, Weight M, Molinari R. Extensive postoperative epidural hematoma after full anticoagulation: case report and review of the literature. J Spinal Cord Med 2007;30:282–7. [10] Doran SE, Papadopoulos SM, Miller LD. Internal fixation of the spine using a braided titanium cable: clinical results and postoperative magnetic resonance imaging. Neurosurgery 1996;38:493–6. [11] Ebraheim NA, Rupp RE, Savolaine ER, Reinke D. Use of titanium implants in pedicular screw fixation. J Spinal Disord 1994;7: 478–86. [12] Ebraheim NA, Savolaine ER, Stitgen SH, Jackson WT. Magnetic resonance imaging after pedicular screw fixation of the spine. Clin Orthop Relat Res 1992;279:133–7. [13] Liljenqvist U, Lepsien U, Hackenberg L, et al. Comparative analysis of pedicle screw and hook instrumentation in posterior correction and fusion of idiopathic thoracic scoliosis. Eur Spine J 2002;11: 336–43. [14] Fisher CG, Sahajpal V, Keynan O, et al. Accuracy and safety of pedicle screw fixation in thoracic spine trauma. J Neurosurg Spine 2006;5:520–6. [15] Kuklo TR, Lenke LG, O’Brien MF, et al. Accuracy and efficacy of thoracic pedicle screws in curves more than 90 degrees. Spine 2005;30:222–6. [16] Haaker RG, Eickhoff U, Schopphoff E, et al. Verification of the position of pedicle screws in lumbar spinal fusion. Eur Spine J 1997;6: 125–8. [17] Reichle E, Morlock M, Sellenschloh K, Eggers C. [Definition of pedicle malposition. Primary stability and loosening characteristics of pedicle screws in relation to position: spongious anchoring, cortical anchoring, perforation and malposition]. Orthopade 2002;31: 402–5. [18] Seller K, Wild A, Urselmann L, Krauspe R. [Prospective screw misplacement analysis after conventional and navigated pedicle screw implantation]. Biomed Tech (Berl) 2005;50:287–92. [19] Brooks D, Eskander M, Balsis S, et al. Imaging assessment of lumbar pedicle screw placement: sensitivity and specificity of plain radiographs and computer axial tomography. Spine 2007;32:1450–3. [20] Learch TJ, Massie JB, Pathria MN, et al. Assessment of pedicle screw placement utilizing conventional radiography and computed tomography: a proposed systematic approach to improve accuracy of interpretation. Spine 2004;29:767–73. [21] Gertzbein SD, Robbins SE. Accuracy of pedicular screw placement in vivo. Spine 1990;15:11–4.

290

S.W. Tohtz et al. / The Spine Journal 10 (2010) 285–290

[22] Suk SI, Kim WJ, Lee SM. Thoracic pedicle screw fixation in spinal deformities: are they really safe? Spine 2001;26:2049–57. [23] Yalcin S, Guven O. Reversible anterior cord syndrome due to penetration of the spinal canal by pedicular screws. Paraplegia 1995;33:423–5. [24] Czervionke LF, Berquist TH. Imaging of the spine. Techniques of MR imaging. Orthop Clin North Am 1997;28:583–616.

[25] Gille O, Jolivet E, Dousset V, et al. Erector spinae muscle changes on magnetic resonance imaging following lumbar surgery through a posterior approach. Spine 2007;32:1236–41. [26] Peng CW, Chou BT, Bendo JA, Spivak JM. Vertebral artery injury in cervical spine surgery—anatomical considerations, management, and preventive measures. Spine J 2009;9:70–6.