Introduction

The reported pedicle screw misplacement in historical spinal literature can be as high as 20–39.8% [1–3], but only a small number leads to complications (neurological, vascular or visceral injuries). But these complications can be potentially life and limb threatening. Computer-assisted navigation allows for simultaneous and multi-planar visualization of spinal anatomy which helps in virtually tracking surgical instruments in relation to displayed anatomy in real time [4]. This has led to its utilisation in pedicle screw placement thus increasing the accuracy of screw positioning in cadavers and patients [5–9].

The question then arises, that does this increased accuracy in screw placement lead to a significant decrease in the complication rates from the misplacement of pedicle screws or increased functional outcomes in terms of better fusion rates, improved pain scores and superior health outcome measures?

The primary research objective for this systematic review was to investigate whether accurate placement of pedicle screws, with the help of navigation, gives benefits in terms of avoidance of neurological complications, improved pain relief, improved fusion rates and better health outcome scores such as Oswestry disability index, SF-36 or SF-12 scores. Our secondary aim was to look at the accuracy of pedicle screw placement with navigation.

Materials and methods

We conducted a systematic review and meta-analysis of the available studies of computer-assisted pedicle screw insertion using the quality of reporting of meta-analysis (QUOROM) guidelines for synthesis of information from the relevant existing literature [10].

Literature identification and inclusion/exclusion criteria

The abstracts and titles of all the articles in: MEDLINE (1950 to July 2009), EMBASE (1974 to July 2009) and CINAHL (1982 to July 2009) were searched via ‘Dialog Datastar’ with the following key words: pedicle screw and navigation. Thesaurus mapping was then used to explode this search with “spine” and combining these searches with the Boolean linkage terms AND and OR to identify relevant publications.

The complete articles identified by the above search methodology were retrieved and assessed against the inclusion/exclusion criteria outlined in Table 1. Additionally, the cross-references in the trials retrieved electronically were hand searched.

Table 1 Inclusion and exclusion criteria
Full size table

Data collection

The data collected from the qualifying articles was: indication for surgery, number of patients, vertebral level(s) instrumented, number of pedicle screws, neurological loss and patient-based outcome measures (fusion rates, Oswestry disability index, SF-36, SF-12, and pain scores) where available. Data was extracted independently by two reviewers, R.V. and S.K. except for articles in German language which was done by K.H. Any discrepancy was resolved mutually and if necessary by third party agreement (A.M.). In view of paucity of randomised controlled trials and heterogeneous nature of trials, we did not use a formal scoring method for assessing quality of studies. Instead, we critically appraised the studies using a checklist designed to assess the methodological quality of both randomised and non-randomised studies [11].

Statistical methodology

Data was pooled separately for studies with control group and for those studies without control group.

Randomised and case–control studies

Summary statistics was created by calculating odds ratio with 95% confidence interval for the various functional outcomes. Heterogeneity was tested using I2 statistics. As there was high heterogeneity in the studies, random effects model was used. Data was entered and analysed in Revman 4.2 [http://www.cochrane.org/ (2008)].

Case series

The case series data was pooled using an inverse variance method weighted for the size of the study. This pooled data was analysed by random effects model.

Results

The electronic search methodology identified 67 possibly relevant publications, while the hand search of the references of these 67 articles identified a further four articles resulting in 71 papers being reviewed. At review, 48 papers were excluded as they did not meet the inclusion criteria (Table 1). Thus, 23 publications [5, 6, 8, 9, 12–30] were analysed in this paper (Table 2; 2 randomised controlled trials, 12 case–control studies and 9 case series). Three studies were in German language and remaining 20 studies were in English.

Table 2 Publications analysed [5, 6, 8, 9, 13–31]
Full size table

These 23 studies in essence include: 719 patients (3,555 pedicle screws inserted with the help of navigation techniques) with an age range of 13–61.2 years and 569 patients (2,437 pedicle screws inserted without the help of navigation techniques) with an age range of 15.4–60.2. The indication for surgery varied widely across the different studies including correction for kyphosis and scoliosis, treatment for metastasis, rheumatoid arthritis and trauma. The surgery was performed at all levels and did not exclude any particular level.

Neurological complication

All of the analysed studies reported about presence or absence of neurological complications as a result of pedicle screw insertion. Navigational techniques were used to insert pedicle screws in 327 patients (9 case series) and 392 patients (2 RCT’s and 12 case–control studies). There were no reported cases of neurological complications in navigational group in either series. On the other hand, conventional pedicle screws were inserted in 569 patients (2 RCT’s, 12 case–control studies) leading to 13 cases of neurological complications (2.3%). The meta-analysis undertaken (Fig. 1) demonstrated relative odds ratio of 0.25 (95% CI of 0.06–1.14) in favour of using navigation for pedicle screw insertion; however, this result was not statistically significant (p = 0.07).

Fig. 1
figure 1

Forest plot showing neurological complications in comparative trials. Individual studies are plotted on y-axis with a summary estimate at the bottom. The point estimates (blue squares) are shown on a line that represents their associated 95% confidence interval. The odds ratio (OR) is shown towards the right of the diagram with the pooled estimate calculated with the random-effects model at the bottom. The pooled estimate suggests that risk of neurological complications decreases with navigation

Full size image

Kotani et al. [29] reported a girl in the non-navigational group, developing a neurological loss after 4 years. She was included in the analysis as the cause was found to be screw perforation and symptoms resolved with screw removal. Seichi et al. [16] reported a case of neurological loss (cervical myelopathy) which was due to tumour re-growth thus this was not included in the analysis.

Fusion rate

Six studies [12, 13, 19, 24, 26, 30] reported follow-up period ranging from 15 to 34 months, while one study followed the patients to clinical and radiological bony fusion [16], but none reported on rate of fusion achieved.

Pain relief and health outcome scores

Amiot et al. reported that two of their patients in the navigation group had dysesthesia in the post-operative period but a conservative approach was followed with symptom resolution at 6 months. While, Ito et al. reported that the Ranawat’s pain score in their ten rheumatoid arthritis patients improved from 1.4 to 1 in the navigation group and from 1.6 to 1.2 in non-navigational group.

No study gave the health outcome scores like the Oswestry disability index or SF-36/12 scores.

Accuracy

All 23 studies (n = 5,992 screws) provided accuracy data. Amiot et al. and Seller et al. used magnetic resonance imaging (MRI) for grading post-operative accuracy while other authors used computerised tomography techniques (CT scan) [5, 6, 8, 9, 12–30].

A total of 93.3% (n/N = 3,316/3,555) of the pedicle screws were inserted accurately with navigational techniques, whereas 84.7% (n/N = 2,064/2,437) were inserted accurately with non-navigational techniques. Meta analysis of the 14 studies with control group (RCT = 2, case–control studies = 12), i.e. 1,838 pedicle screws from the navigation group and 2,437 from the non-navigational group showed a significant advantage (p < 0.00001) of navigation over non-navigation (conventional) pedicle screw insertion with a relative risk of 1.12, with a 95% CI 1.09, 1.15 (see Fig. 2). Pooled data from the nine case series (1,717 screws, 48%) that used navigational techniques also showed accurate placement of pedicle screws (risk ratio was 0.92 with 95% CI 0.88, 0.96).

Fig. 2
figure 2

Forest plot showing accuracy of placement of pedicle screws in comparative trials. The relative risk (RR) is shown towards the right of the diagram with the pooled estimate calculated with the random-effects model at the bottom. The diagrammatic pooled estimate (diamond) and numerical relative risk both shows that navigation helps in more accurate placement of pedicle screws

Full size image

Discussion

Pedicle screws have become the favoured method of posterior spinal fixation since being popularised by Roy-Camille [31]. Biomechanical studies have shown that pedicle screws provide increased rigidity and construct stiffness compared to other posterior fixation techniques [32, 33]. It is a technically demanding procedure and studies have shown an alarmingly high rate of screw misplacement varying between 21 and 39.8% with conventional insertion techniques [1–3]. The misplaced screws can cause three types of complications, namely neurological, vascular and mechanical.

Neurological complication while inserting pedicle screws is a rare but serious complication, which can be avoided by careful planning of the surgery. Theoretically, image-guided navigation techniques can help the surgeon in preventing the complication by providing accurate anatomic guidance for the procedure. However, Schulze et al. [34] have argued that experienced surgeons can accurately place pedicle screws in 80% of cases with conventional techniques and neurological symptoms are rarely affected by an inaccurate pedicle screw even if the penetration of the pedicle wall is more than 6 mm. In our meta-analysis, pedicle screw insertion by navigational techniques have caused fewer incidences of neurological complications than conventional methods though this was not statistically significant (p = 0.07). We think this could be due to lack of big randomised trials as the incidence of neurological complications is low and none of our studies had big numbers required for narrow confidence intervals.

We were unable to make conclusions regarding the fusion rates after computer-assisted surgery because of the paucity of information provided by many authors. None of the studies in our meta-analysis reported any vascular complications and none of them used patient-based outcome measures like Oswestry disability index, SF-36 or SF-12 scores.

Limitations

These results must be interpreted with caution as there were various potential sources of heterogeneity. Insertion of pedicle screws are complex interventions and it is difficult to avoid bias in comparison groups. Our a priori hypothesis for heterogeneity included diverse nature of studies, different indications for surgery and varying complexity of surgery as well as different levels of spine. Randomised controlled trials provide the strongest evidence for meta-analysis. Our study had only 2 RCT’s and the rest were controlled trials, some of which were retrospective. Inherent biases and confounding factors in non-RCT studies sometimes are unavoidable. Meta-analysis and systematic reviews are not without its flaws, but it is not our intention to discuss them here. For a review of the criticisms, the reader is directed to papers by Greenland [35, 36].

Conclusion

Does computer navigation provide better clinical outcome than conventional techniques? We did not find any statistically significant difference in the available clinical outcomes between two techniques. Current evidence does not favour computer-assisted navigation over conventional techniques.

In this context it is useful to remember that ‘absence of evidence is not evidence of absence’ [37]. Further research in this area should include randomised controlled trials with well-planned methodology to limit bias and report on validated patient-based outcome measures.