Navigation & Robotics in Spine Surgery

The metric. Pedicle-screw accuracy is the central measure, graded by the Gertzbein-Robbins classification (set out in full below). Both navigation and robotic assistance improve accuracy over conventional freehand and fluoroscopy-guided insertion.

Navigation. Intraoperative imaging is registered to the patient so that instruments are tracked in real time against the image, letting the surgeon plan and follow a pedicle-screw trajectory. The most powerful form is 3D imaging from an intraoperative CT or O-arm; fluoroscopy-based 2D-3D systems are the alternative. It gives good accuracy with relatively shorter operative times than robotics.

Robotics. Systems such as Mazor X, ROSA, ExcelsiusGPS and Cirq position a guide or arm along a pre-planned trajectory for screw insertion. Meta-analyses show grade A/B accuracy of around 98%, with no single platform clearly superior to the others, and greater accuracy than traditional methods. The cost is generally longer operative time, especially with setup and registration, and higher cost.

What else they buy. Both generally bring fewer intraoperative revisions and reoperations than freehand or fluoroscopic insertion, although the reoperation benefit rests on low to very low certainty evidence (see the outcomes section below). Radiation to the surgeon and theatre staff falls because the surgeon is away from the beam during navigated or robotic insertion. Intraoperative CT, however, delivers a dose to the patient that must be justified and optimised, especially in children.

Choosing between them. Robotics tends to give the highest accuracy, which is valuable in complex and deformity cases, but takes longer; navigation gives good accuracy efficiently. The choice balances precision against time and cost, and freehand or fluoroscopy remains valid for standard cases. In deformity surgery there is generally no significant difference between navigation and robotics in blood loss, curve correction or hospital stay.

The price. High capital and per-case cost, a real learning curve and added setup and registration time are part of the decision.

The technical failures. Each is a reason to verify the guidance rather than trust it blindly.

• Registration error - the navigation image no longer matches the patient after reference-frame movement, inter-segmental motion between registration and screw insertion, or drift, so the on-screen trajectory no longer corresponds to the real anatomy
• Line of sight - optical tracking needs an unobstructed line of sight between the camera and the trackers, and it can be lost
• Skive - the drill or tap can deflect off a sloped pedicle or facet despite correct planning

Every accuracy figure quoted for navigation and robotics is a Gertzbein-Robbins grade, so the whole scale is worth knowing, not just its acceptable end.

• Grade A - fully contained within the pedicle, no cortical breach
• Grade B - cortical breach of less than 2 mm
• Grade C - breach of 2 to 4 mm
• Grade D - breach of 4 to 6 mm
• Grade E - breach of 6 mm or more

The acceptable threshold. Grades A and B together are considered clinically acceptable, because a breach under 2 mm is within the safety margin. That is why the literature reports combined grade A/B rates.

Direction matters too. A medial breach threatens the cord and nerve root and is the dangerous one, and an inferior breach risks the exiting root. A lateral breach is usually safer but sacrifices fixation purchase, so accuracy means a screw that is contained or safely directed, not only a grade.

View larger

The dynamic reference frame. The array is a tracked marker clamped rigidly to a fixed bony landmark: a spinous process, or the iliac crest for lumbosacral and pelvic work. Every instrument position is computed relative to this frame, so if it is bumped or shifts, the whole on-screen map is wrong even though it still looks plausible.

Matching the image to the patient. This is the registration method proper.

• Automatic image-based registration from an intraoperative 3D scan (CT or O-arm) - the most powerful: it co-registers the just-acquired volume to the tracked frame with no manual point-picking, and is fast and accurate
• Paired-point plus surface matching to a preoperative CT - anatomical landmarks are picked, then the surface is mapped; more error-prone
• 2D-3D fluoroscopy-based registration

Why accuracy decays. The spine is not a rigid body. Registration is exact at the reference level but degrades at levels further away because of inter-segmental motion. So verify or re-register per region, re-spin or re-register after a major correction manoeuvre or any bump, and keep the array out of the working field. Always confirm registration against a known landmark before trusting the guidance. General CT and fluoroscopy physics and dose are covered in CT Physics, Radiation Dose and Hounsfield Units and Radiation Safety in Orthopaedics.

"Verify against the anatomy and do not trust the screen" is sound advice and useless without the checks themselves. There are four, and they are independent of one another, which is the point: each fails in a different way, so agreement between them is what provides confidence.

The triggered-EMG trap. The intuition runs the wrong way: a low threshold is the worrying result, because current is escaping through thin or breached cortex to reach the nerve. A high threshold means intact bone is insulating the screw. Modalities, alarm criteria and the rest of the monitoring picture are in Spinal Cord Monitoring and IONM.

How the guidance fails. Navigation and robotics fail intraoperatively in ways that are recoverable if anticipated: the array is knocked, line of sight is lost, registration will not hold at the caudal levels, the robot arm skives off a sloped facet, or the scan is degraded by artefact from existing metalwork.

The bail-out. The response is a sequence, not a crisis.

1. Re-register and re-verify against a landmark
2. Re-spin the intraoperative scan if the geometry has changed
3. Move the reference array closer to the level being instrumented
4. If accuracy still cannot be confirmed, convert to freehand or fluoroscopically guided placement with direct palpation

Any surgeon using navigation must remain able to place a screw without it. Losing the technology mid-case is an inconvenience for a surgeon who can do that and a genuine emergency for one who cannot, which is the strongest practical argument for continuing to teach and practise freehand technique alongside the machines.

An examiner who has heard the accuracy figures will ask the question that matters: a screw that is 2 mm better placed - does the patient notice? The evidence for accuracy is consistent. The evidence that it changes how patients do is much weaker, and a candidate who claims otherwise is easy to dismantle.

Why the two questions come apart. Most malpositioned pedicle screws cause no symptoms whatsoever: a lateral breach may sacrifice purchase without ever troubling the patient, and even a modest medial breach is frequently tolerated. Symptomatic malposition requiring revision is uncommon, so the event that better accuracy would prevent is rare, and a technology that reduces a rare event has to be studied in very large numbers before any effect on that event becomes visible. Accuracy is a surrogate outcome: easy to measure, measured in every study, and only loosely coupled to what the patient experiences.

What the appraised evidence says. An umbrella review took 22 published meta-analyses of robot-assisted versus freehand technique and assessed them formally with AMSTAR-2 and GRADE, across 14 associations including "perfect" and "clinically acceptable" screw placement, complications, facet joint violation and radiation. Robot assistance came out favourably, but the certainty of the evidence for the majority of those associations was rated low to very low, and the authors concluded that the findings need cautious interpretation before being applied in practice. Much of the underlying literature is observational, subject to selection of straightforward cases into the new technology and of difficult ones away from it, and reported by enthusiasts during their own learning curves.

The randomised evidence is still arriving. The ROBARTHRODESE trial randomises 300 patients to robot-assisted versus fluoroscopic freehand thoracolumbar fusion, and its primary outcome is the proportion of Grade A screws, with disability score, pain, complications, reoperation, radiation and a cost-utility analysis as secondary outcomes. The surrogate is still the primary endpoint even in the trials designed to settle the question.

How to say this in a viva without sounding either credulous or dismissive. The technology reliably does what it claims, placing screws more accurately and taking the surgeon out of the radiation beam, and both are worth having. Its value is greatest where the anatomy makes freehand genuinely hazardous: severe deformity, revision with distorted landmarks, high thoracic and cervicothoracic levels, sacropelvic fixation, obesity, and percutaneous minimally invasive work where landmarks are not exposed at all.

What it has not yet been shown to do is make an average lumbar fusion turn out better. The case for buying one rests as much on radiation protection and on capability in complex cases as on outcomes data that does not yet exist.