Arthrodesis of the spine by the posterior approach is the most common surgical treatment for many spinal deformities of childhood and adolescence. Here, posterior spinal fusion for a  variety of conditions, emphasizing surgical management of idiopathic scoliosis is discussed. Operative treatment of congenital scoliosis and neuromuscular scoliosis is also briefly discussed.  Preoperative planning, intraoperative control of blood loss, and postoperative management are discussed as well as the important role of the anaesthesiologist.

Posterior spinal fusion for scoliosis was first described by Hibbs in 1911. From his experience with tuberculosis of the spine, he surmised that arthrodesis of the spine would probably prevent progression of idiopathic scoliosis. Hibbs' technique emphasized meticulous dissection. careful hemostasis, and prolonged postoperative immobilization. No autogenous iliac crest bone graft was used nor were the facet joints violated.

Over the next 40 years, arthrodesis of the spine for scoliosis met with limited success. There was an unacceptably high rate of failure of fusion, and the amount of correction that could be obtained was limited. The use of additional bone graft from the iliac crest was helpful. Moe emphasized careful excision of the facet joints to improve the chance for a successful arthrodesis.

In the early 1960s, Harrington reported on the use of distraction and compression rods for internal stabilization of spinal deformities and ushered in the modern era of spinal fusion for scoliosis. The past 25 years have seen tremendous growth in the number of devices available for internal fixation of the spine. Today the scoliosis surgeon must not only master the technique of exposure and arthrodesis of the spine, but must also become familiar with the various types of internal fixation and how to use them.

The major goals in spinal fusion for scoliosis are to prevent progression of the curve to the point that the patient is disabled by cardiorespiratory compromise, pain, or neurological deficit and to correct the deformity. The psychological effects of a large spinal deformity are incompletely understood, but may be very important to the patient. Partial correction of the scoliosis is possible, but patients must be informed that their spine will never be normal, because correction of the curve and arrest of progression are obtained at the cost of reduced spinal motion. Because the operation is performed to prevent progression, the curve should be at significant risk for progression. Factors used to assess the risk of progression include the age and maturity of the patient, the magnitude of the curve, and the location of the curve.

The Risser method of grading using the ossification of the iliac apophysis is helpful in assessing the degree of maturity and determining the risk of scoliosis progression. Curves in patients of Risser stage 3 or higher are much less likely to progress than curves in patients of Risser stage 2 or less. In idiopathic scoliosis, curves below 40° Cobb angle are unlikely to progress after skeletal maturity and seldom require fusion . Curves above 50° Cobb are more likely to progress after skeletal maturity, particularly those in the thoracic spine.

Curves between 40° and 50° Cobb represent a gray zone where observation may be indicated for skeletally mature patients in good alignment, and arthrodesis should be considered in skeletally immature patients who are at risk for further progression during their remaining growth. Thoracic and thoracolumbar curves are more likely to progress than lumbar curves. The deformity that results from scoliosis can vary greatly, even for curves of the same magnitude. Idiopathic scoliosis is a three-dimensional spinal deformity. Alterations in spinal alignment such as decreased thoracic kyphosis and increased vertebral rotation contribute to the unsightly "rib hump", but may not be apparent on the anteroposterior radiograph. The scoliosis may result in the head not being centered over the pelvis. A slender patient's deformity may be more apparent than that of a patient who is obese.

Although the decision for surgery is based mainly on the spinal deformity in the frontal plane, a lateral radiograph of the spine should be taken before surgery to assess the sagittal alignment of the spine and to rule out concomitant problems such as spondylolisthesis. To aid in selection of fusion levels, the flexibility of the spine should be evaluated. Radiographs can be obtained with the patient lying down while someone attempts to reduce the deformity with manual pressure or traction, or with the patient standing with instructions to lean as far to each side as possible.

In curves of large magnitude (more than 90° Cobb), a two-stage arthrodesis of the spine may be advisable. The first stage consists of an anterior release of the spine, followed possibly by halo traction. Posterior spinal fusion is performed a week or two later. Careful attention has to be paid to the patient's nutritional status between the two operations. Hyperalimentation should be strongly considered to minimize the catabolic effects of the first procedure and to limit complications.

Before surgery, the levels of fusion and instrumentation sites should be selected. Deciding how much to fuse depends on curve location and flexibility of the compensatory curves. Too short a fusion may result in further progression of the deformity, while too long a fusion requires a larger operation and costs the patient motion segments. There are several ways to determine the extent of a planned spinal fusion. A rule of thumb is to fuse one or two vertebrae above and below any structural curves. A structural curve is defined as one that does not reverse on bending films and in which the spinous processes are rotated toward the midline.

In addition, instrumentation should be placed in such a way to center the head over the pelvis. The lower hook should be seated within Harrington's stable zone, which is defined by vertical lines extending up from the sacroiliac joints. King et.al emphasized the importance of fusing from neutrally rotated vertebra to neutrally rotated vertebra, that is, a vertebra whose spinous process is equidistant from the two pedicles on the anteroposterior radiograph. He also pointed out that many compensatory lumbar curves do not require fusion if they are more flexible than the accompanying structural thoracic curve.

Fusion to the last lumbar vertebra (usually the fifth) has been associated with a high incidence of back pain and should be avoided.

Substantial blood loss should be anticipated from the extensive exposure required, making blood replacement almost always necessary. To minimize the risk of transfusion-acquired illnesses, it is preferable to have patients donate their own blood preoperatively for retransfusion later. Patients are allowed to donate 1 unit per week up to 10 days before their surgery. With new preservatives, this allows them to donate 3 to 4 units in the month before surgery. Patients are instructed to take ferrous sulfate, and the blood bank will continue to draw blood for autotransfusion as long as the patient's hematocrit remains above 30%.

With the blood loss and fluid shifts that inevitably occur during scoliosis surgery, close monitoring of the patient is essential. A foley catheter should be inserted to monitor urine output. The state of the circulatory system can be measured in several ways. An arterial catheter can measure peripheral blood pressure. A central venous pressure line can be inserted to administer large amounts of fluids rapidly. The reliability of monitoring the central venous pressure in the prone position is questionable. The most accurate way to assess the circulatory system is with a right atrial (Swan-Ganz) catheter that directly monitors cardiac output. A Swan-Ganz catheter is more difficult to insert than a central venous or arterial catheter. A possible advantage of a Swan-Ganz line is the ability to aspirate air embolism, if it occurs, from the right side of the heart. A Doppler stethoscope taped to the chest can detect this rare but catastrophic complication.

A major concern in scoliosis surgery is protection of neurological function. Earlier recognition of neurologic compromise allows a more rapid response. which may increase the chance of neurological recovery. Intraoperative assessment of spinal cord function has become routine for most scoliosis surgeons. Spinal cord function can be assessed intraoperatively by the wake-up test, the clonus test, or somatosensory-evoked potentials (SSEP). The wake-up test is performed by bringing the patient up to a light level of anesthesia. The patient is instructed to squeeze the anesthetist's hand to make certain the patient is awake enough to follow commands. The patient is next told to move the feet. Once this has been observed, me anesthesia is deepened and the procedure continue, The wake-up test requires some comprehension on the patient's part. although most do not recall the episode later. Rehearsal with the patient before surgery enhances its reliability.

Close cooperation with the anesthesiologist is crucial. Gradual lightening of the level of anesthesia cal! allow rapid performance of the wake-up test when the surgeon wants it. Care must be taken to avoid complete reversal of anesthesia, or the patient's blood pressure may rise excessively and the patient's movements may disengage the fixation and dislodge the endotracheal tube.

The clonus test is performed on a patient at the same level of anesthesia as a wake-up test, but requires no patient cooperation. Instead of commanding the patient to move the feet, an examiner rapidly dorsiflexes the ankles. The production of sustained clonus indicates preservation of spinal cord function.

Many centers perform intraoperative assessment of spinal cord function by SSEPs. Stimulating electrodes are placed over the posterior tibial and popliteal nerves. Electrodes on the skull record the signals transmitted through the spinal cord from the legs. SSEP signals of diminished amplitude or increased latency indicate possible problems with spinal cord function. SSEPs monitor only posterior column (afferent nerve) function. and there have been cases of paralysis without SSEP change. The use of SSEPs places additional demands on the anesthesiologist. Inhalation agents such as halothane can not be used in large dosages. Extra care must be taken to maintain the patient's core temperature, because hypothermia can alter the SSEP signals. Wrapping the extremities in blankets before surgery, and heating and humidifying the inspired anesthetic gases, can help reduce intraoperative heat loss.

Careful positioning of the patient can also help reduce blood loss. Placing the patient prone on a Relton-Hall frame allows the abdomen to hang free, taking pressure off the vena cava. This reduces the blood flow through Batson's plexus, which surrounds the spinal cord. A spinal fusion is a relatively long procedure. The patient will be in the prone position on the Relton-Hall frame for 4-6 hours. The legs should be wrapped in elastic bandages to help prevent thrombus formation and to reduce heat loss. Bony prominences, in particular the patellae and iliac crests, should be carefully padded. In females, care must be taken in positioning on the frame to ensure that the breasts are not under too much pressure. The arms should be placed with the shoulders in less than 90° abduction or brachial plexus palsies can result. An x-ray plate should be placed beneath the patient to allow intraoperative radiographic localization of the levels exposed.

The use of internal fixation as an adjunct to a well-performed arthrodesis of the spine has greatly improved the results for surgical management of scoliosis. With spinal instrumentation, greater correction can be obtained; the rate of pseudoarthrosis is reduced, and patients can be mobilized more quickly. Today's scoliosis surgeon is faced with a bewildering array of devices for internal fixation of the spine. Each type of instrumentation has its own advantages and disadvantages, and each surgeon needs to be familiar with his own abilities to use the different forms of instrumentation. Use of internal fixation devices facilitates correction of deformity, but may decrease the amount of bone exposed and available for arthrodesis. The benefit of slightly increased correction must be weighed against the difficulty of application of a specific spinal instrumentation system and the risk of complications. If a major problem is encountered intraoperatively, the ultimate fallback is to use no instrumentation and obtain correction with a Risser localizer cast postoperatively.

Unfortunately, the ideal spinal instrumentation system does not exist. The ideal system would have a high degree of safety and reliability with a minimum risk of neurological injury. The rate of complications such as failure of fixation and implant breakage would be low. The system would be strong enough to resist loads and moments from all directions, perhaps without external support. It would be easy to apply, with little increase in operative time over that needed to perform an arthrodesis. The ideal system would restore normal spinal contours and not create new deformities.

Each of the currently available spinal instrumentation systems achieves some of these criteria. None meets them all. No one device is the best choice for every surgeon to use on every patient. Surgeons vary in technical ability and experience, and patients vary in their needs and desires. In interviews with patients, parents, and physicians, lack of postoperative immobilization was not regarded to be as important as the avoidance of neurologic complications.
We briefly review here the various devices commonly in use to augment posterior spinal fusion in the management of idiopathic scoliosis. These include the Harrington distraction and compression rod system, the Luque method of sublaminar wire fixation, the Wisconsin method of fixation with spinous process wires, the Cotrel-Dubousset system. Recent modification improvements made in these systems are also discussed.

The spinal implants developed by Dr. Paul R. Harrington. The instrumentation originally consisted of a distraction rod applied to the concave side of the curve and a compression rod applied to the convex side. After installation of Harrington rods, a generous area of lamina remains accessible for decortication and bone grafting. An average curve correction of 50% can be obtained using Harrington rod in adolescent idiopathic scoliosis. Several factors limit further correction. Excessive distraction may be associated with neurological injury, presumably from stretch-induced spinal cord ischemia. Prompt removal of distraction will sometimes reverse the neurological injury. Intraoperative assessment of spinal cord function by SSEP, the wake-up test, or the clonus test may help lessen the risks this potentially catastrophic complication.

A second limiting factor to curve correction by Harrington rod is the decreased efficiency of axial distraction - the curve magnitude reduces. Axial distraction corrects a spinal curvature by applying a force parallel to the spine acting across a moment arm that runs perpendicularly to the apex of the curve. As the curve is reduced, this moment arm is decreased, so greater applied force results in less correction. With Harrington distraction rods, all the distraction force is applied only at the two laminae where the hooks are seated. If the load exceeds the strength of the lamina, fracture and loss of correction can result. Because the thoracic laminae are thinner failure is more likely to occur at the upper hook site.

To avoid overdistraction, some surgeons incorporate strain gauges into their distraction devices. Other rely on "feel" for when to stop distracting. Studies of instrumented spines have demonstrated that the load on the distraction rod diminishes in a few minutes because of the viscoelasticity of the spine's soft tissues (creep). Repeated distraction over time can gradually overcome much of the viscoelasticity and allow further correction. Repeated distraction of the Harrington rod can overcome the creep to some extent, but a problem arises; as the spine is lengthened, the ratcheted portion of the rod below the hook becomes longer. This increases the stress on the junction of the ratcheted (thin) portion of the rod and the smooth (thick) portion of the rod, and increases the risk of rod fracture at the junction. Distraction rods with short ratcheted segments are manufactured to minimize the risk of rod fracture, but these limit the ability to repeatedly distract the spine.

Distraction of the spine can also be done with an outrigger device attached to the hooks. By using the clothespin-like hook clamps devised by Zielke, it is possible to keep spine closer to the final length while placing a distraction rod that has a short ratcheted portion.

One method of reducing the risk of lamina fracture is hare the load between two adjacent laminae. With regular ratchet end hooks this is difficult, because the sagittal curve of the spine may not allow placement of the hooks in a line straight enough to allow passage of distraction rod through two hooks. Bobechko devised hook with a sliding barrel that allows placement of two hooks on the ratchet end of the rod. A special spreader is used to load both hooks simultaneously. The sliding barrel hook is taller than a regular hook, and may be prominent or even subcutaneous in thin individuals. This construct uses a straight distraction rod that reduces not only scoliosis but also the normal sagittal plane curves.

The Harrington distraction rod elongates the spine. This elongation results in reduction of spinal curves in both the coronal and sagittal planes. The coronal plane curve, is, scoliosis, is pathological. The sagittal plane curves, that is, thoracic kyphosis and lumbar lordosis, are physiological. If the spine is instrumented to the lower lumbar region with a straight distraction rod, the normal lumbar lordosis is reduced and a "flatback" syndrome can result. To walk upright, affected patients must either hyperextend their hips or walk in a crouched position. Neither is desirable, and both are uncomfortable and unsightly.

To overcome this problem, Moe modified the Harrington rod by making the inferior end of the rod, and the corresponding hole in the lower hook, square shaped. This allows the rod to be contoured to preserve lumbar lordosis while preventing unwanted rotation of the rod. While the use of contoured rods diminishes the loss of sagittal plane curves, it has not completely eliminated this problem. Contoured rods are weakened and more likely to break, and the mechanical efficiency of the rod for distraction is reduced. Square-ended contoured rods are also more difficult to install. A special pusher or clamp is very helpful, because rotation must be controlled to seat the rod in the square-holed Moe hook.

The strength of a Harrington rod construct depends on how well the hooks are attached to the lamina. In axial loading, the limiting factor is the strength of the lamina in which the hooks are seated. The spine fails before the Harrington rod breaks. With bending and particularly twisting of the trunk, hooks can be dislodged fairly easily, resulting in loss of fixation. Because of the risk of dislodgement, most surgeons use some form of external immobilization, such as a cast or brace, after Harrington instrumentation.

Several hook modifications have been developed to help prevent accidental dislodgement. Zielke devised a ratchet-end hook with a bifid shoe designed to fit tightly around the pedicle of the thoracic vertebra. A ratchet-end hook with a sharp rib in the shoe can also be used. The rib helps prevent lateral displacement of the hook, but can weaken the bone and lead to lamina fracture. Leatherman designed a series of collar-end hooks with extended shoes that help improve seating of the hook in a spine with increased lumbar lordosis. One problem is that if a Leatherman hook shifts or tilts, it can narrow the spinal canal. Andre hooks are angled so that the rod insertion site is no longer parallel to the shoe. This hook conforms better to the lamina and is less likely to shift or tilt.

For instrumentation to the pelvis, a variety of hooks have been designed to attach to the sacrum. While the Moe alar hook is best known, its mouth is so wide that the risk of accidental dislodgement is increased in smaller patients. Leatherman hooks with a smaller mouth may conform better to the ala in small patients and may fit better in some cases. Another method of instrumenting to the pelvis is placement of a threaded rod through both posterior iliac crests just above the sacrum. The distraction rod can be attached by either an eyelet threaded on the sacral rod or a special collar-end hook with threads on its shoe to interlock with the sacral rod. Zielke developed a hinged rod that is driven into the ilia bilaterally. It is less prominent, an advantage in thin patients.

Prominence of the hardware beneath the skin poses a serious problem in young children or thin patients, particularly when the spine is instrumented into the upper thoracic region (T3 and above). A simple solution is to invert the rod and hooks and place the smaller collar-end hook cephalad. The ratchet-end hook can then be placed caudad where more muscle is available to cover it. Pediatric hooks with smaller shoes and bodies are also available. A threaded rod can be used for distraction in smaller patients, although it is thinner than the standard distraction rod and not as strong.

The Harrington compression rod entails placement of multiple hooks linked on a threaded rod around either the transverse processes or laminae. Nuts adjacent to each hook are moved along the rod by a wrench to apply the compressive force. While the threaded rod is flexible, proper placement of the hooks can be difficult and time consuming.

Keene et. al. modified the compression hooks to make them easier to apply. Instead of threading all the hooks on the rod before installation, special hooks with an upward-opening groove are individually placed. The rod is pushed into the hook and a bushing is slid into the hook around the rod, locking the rod and hook together. The wrench used to tighten the nuts generates a large torque without the sensory feedback of the distraction rod spreader; thus, overtightening the nuts can lead to lamina fracture.

The compression rod is less frequently used because many surgeons report good results with just the distraction rod. Correction is not increased substantially with the use of the compression rod, but the stability of the instrumented spine (the ability to resist external loads) may be increased by transverse loading linking the compression rod to the distraction rod. The compression rod tends to increase lordosis in the instrumented segment because it is applied posteriorly on the spine. Because many patients with thoracic scoliosis are already hypokyphotic, an additional flattening of the thoracic spine is undesirable. The compression rod has been recommended to help preserve lumbar lordosis in fusions of the lumbar spine.

In summary, how closely does Harrington instrumentation approach our ideal? It is reasonably safe, if excessive distraction is avoided. In most forms, external support. is advisable postoperatively. It is easy to install. It corrects scoliosis, but care must be taken to prevent unwanted decreases in sagittal plane curves. The use of Harrington rod is vanishing at the present time.

In the mid 1970s, Dr. Eduardo Luque of Mexico City developed a different approach to spinal instrumentation to meet the needs of his patients. Faced with a large number of patients with paralytic scoliosis (from polio) and a climate that made prolonged postoperative immobilization impractical, Dr. Luque developed what he called segmental spinal instrumentation (SSI). SSI consists of 2 contoured smooth rods fixed to the spine by multiple sublaminar wires. Rods of either 3/16- or 1/4-in. diameter are available.

Passage of sublaminar wires requires laminotomies with excision of the ligamentum flavum at each instrumented level. This increases operative time by at least 45 minutes when compared with Harrington rod instrumentation. Luque recommended bending the rods preoperatively, but many surgeons contour the rods intraoperatively, which can be very time consuming, particularly with the stiff 1/4-in. rod. Use of a malleable aluminium rod as a template is helpful. The placement of hardware at every level limits the amount of exposed bone available for fusion.

The Luque system corrects by transverse loading. Its moment arm extends perpendicularly from the apex to the end of the curve. As the curve is reduced, the spine elongates and the moment arm increases. The increasing corrective force as the curve is reduced means that overcorrection with an increased risk of neurological injury is possible. To reduce the risk of overcorrection, the rods should be contoured to include the residual scoliosis achieved on preoperative bending films. A Luque-instrumented spine resists bending and twisting well, but may settle with axial loading.

Because the rods can be contoured in all planes, it is possible to correct scoliosis and preserve sagittal plane curves while the "L" shape helps prevent rotation. The use of multiple sublaminar wires allows correction of thoracic lordosis by pulling the spine to a contoured rod.

Scoliosis correction equivalent to that achieved by Harrington rods is possible. Postoperative immobilization is thought by some to be unnecessary. The resulting construct is so strong that rod breakage from fatigue may not occur for at least several years post surgery. If the rods break, they are most likely to do so at the apex of their contour.

The major drawback of the Luque system is its use of sublaminar wires for fixation. While the spinal cord does not occupy the entire width of the spinal canal, the passage of curved wires can impinge on the dura and spinal cord. Neurological complications ranging from hypoesthesia to complete paralysis have been reported. The incidence of neurological complications has been reported to be as high as 17%. Wilber et al. showed there was a learning curve for many surgeons and, as experience increased, the incidence of neurological complications decreased.

Broken wires have been reported to cause neurological injury by "springing" more deeply into the canal, similar to opening a safety pin. Removal of broken wires may be hazardous as well, because the wire may spring further into the canal as it is extracted. This is perhaps more of a concern in the acute setting. Later the formation of scar tissue around the wire may offer some protection to the dura during wire removal.

Because the spine is viscoelastic, the initial tension in the wires holding the rod diminishes with time. If not properly secured, the rod can migrate or spin. If the rod migrates, it can slip off some wires and become more prominent. If the rod spins, the contour of the rod in the sagittal plane can add to the scoliosis and result in loss of correction. To help prevent rod migration and spin, McCarthy devised a collar that locks to the rod with a bolt. The collar has a hole for passage of a wire to secure it to the spine.

Fixation of Luque rods to the pelvis can be accomplished by putting the transverse portion of the L through a hole in the ilium, not unlike a sacral bar. More secure pelvic fixation can be achieved by the Galveston technique, that is, driving the rod into the pelvis between the inner and outer tables just above the sciatic notch. Contouring the rod to accomplish this is very demanding, but can be done preoperatively.

Moseley has developed a special rod for use in paralytic scoliosis associated with muscular dystrophy. Spin and migration are prevented with his "unit rod" because the entire construct is one long precontoured rod with a hair-pin turn at the top. Use of a closed loop, either a commercially available welded ring or a rod bent into a rectangle, can also help prevent rod migration and spin. Contouring these rods to accommodate the residual three-dimensional spine deformity after correction of scoliosis may be difficult.

In summary, the Luque system is strong, although many surgeons still prefer to use some form of brace for postoperative immobilization. It can correct scoliosis and preserve and even improve sagittal plane curves. In most hands there is a greater risk of neurological injury, which is a major drawback. The risk of neurological injury appears to be at least partially related to surgical experience with the technique. The passage of sublaminar wires and the need to contour the rod substantially increases operative time when compared with the time needed to install a Harrington rod. While very useful for paralytic scoliosis, the use of Luque rods for routine adolescent idiopathic scoliosis is difficult to recommend because of the increased risk of neurological injury.

After Dr. Luque popularized the use of sublaminar wires, many surgeons began securing Harrington distraction rods to the spine with sublaminar wires (Harri-Luque or "Tex¬Mex" technique). This was attractive because both axial and transverse loading were used to correct the scoliosis, and the resulting construct was very strong. The risk of neurological injury remained, not only from the sublaminar wires but also from the possibility of the hooks being pulled horizontally further into the spinal canal as the Harrington rod was loaded transversely and pulled to the spine.

Drummond at the University of Wisconsin developed an alternate technique of segmental fixation. With Wisconsin system, the risks from sublaminar passage of wires are avoided. Wires are passed through a hole made in the base of the spinous process. Each wire is a closed loop that contains a steel button. After the doubled wire is passed, the button is pulled against the spinous process, giving Wisconsin wires a pullout strength almost equal to that of sublaminar wires.
Wisconsin wires were used to secure a contoured Harrington distraction rod on the concave side of the curve and a contoured Luque rod with L bends at both ends on the convex side. While obviously more time consuming than placing a standard Harrington rod, installation of the Wisconsin system is easier, quicker, and neurologically safer than the Luque system. However, Wisconsin instrumentation of double-curve-pattern scoliosis can be challenging, because the recommended technique is to overlap the supporting Luque rods for the two curves by several levels. Leaving the L bend off the overlapping end of one or both Luque rods makes it easier to place them, but increases the potential for rod migration. As with the Luque technique, the multiple rods used by the Wisconsin system limit the area available for bone graft placement.

The Wisconsin system allows for both axial and transverse loading to correct the scoliosis. Theoretically this is the ideal method of correction. Because the rods are contoured and segmentally fixed, and sagittal plane curves can be preserved. The resulting construct is claimed to need minimum postoperative immobilization, perhaps just a corset.

In summary, in idiopathic scoliosis, the Wisconsin system is as safe as a Harrington rod and nearly as strong as a Luque construct. It can be installed relatively easily. By contouring the rods and fixing them to the spine in a segmental fashion, sagittal curves can be preserved while correcting the scoliosis. The Wisconsin system was the instrumentation of choice for routine idiopathic scoliosis in some institutions.

An addition to the armamentarium of the spinal surgeon is Cotrel-Dubousset (C-D) instrumentation. The rod for the C-D system is knurled, allowing hooks to be placed pointing in either direction (cephalad and caudad) with any degree of rotation. This ability to place -hooks in either direction anywhere on the rod allows compression and distraction of different portions of the spine with a single rod. The hooks for the C-D system are available in several shoe shapes, both open and closed. Hooks at the ends of the rod are closed on top and the rod has to be threaded through them. Additional hooks along the rod are open on top and the rod is dropped into them. The open hooks are locked to the rod by hook blockers that are similar in function to the bushings of the Wisconsin compression system. The C-D rod can be rotated while maintaining its attachment to the spine, theoretically allowing correction of rotational deformities and reducing the rib prominence. Two rods are inserted and linked with an adjustable transverse traction rod. Postoperative immobilization is claimed to be unnecessary.

Careful preoperative planning is crucial with the Cotrel-Dubousset (C-D) apparatus. Bending radiographs are needed to plan the multiple hook placement sites. Usually four hooks are placed along the concave side of the curve and four along the convex side. Hook site preparation is similar to that required for Harrington hooks. Fitting the contoured rod to the multiple hooks can be very demanding. Special instruments are available to help place the rod in the hooks. While helpful, these instruments can exert enough force to pull the hook out by fracturing the lamina. Decortication is recommended before instrument insertion because it is difficult after C-D instrumentation is inserted. Given the time it takes to install a C­D rod, this may increase blood loss. C-D instrumentation is much more difficult to install and should be learned from a surgeon experienced in its use. Its theoretical ability to correct the spinal deformity in all planes makes it an exciting new device. Early reports on the C-D system are encouraging. The quest for the ideal spinal instrumentation system will probably never end. Further improvements of the current systems will inevitably appear, new concepts will be developed, and spinal surgeons will continue to be faced with the need to compare the advantages and disadvantages of different systems. Each surgeon should select what will work best in his or her hands for each patient.

Control of blood loss during scoliosis surgery has always been a primary concern for surgeons. Excessive blood loss slows down the operation and increases the risks for the patient. As Hibbs noted in an early report on spinal fusion for scoliosis, only in an operative field that is free from hemorrhage can the operator see to exercise the care necessary for thorough work. Excessive blood loss requires replacement, and the risk of diseases transmissible by blood transfusion such as AIDS and hepatitis, makes it important to minimize the need for allogeneic blood transfusions.

By combining surgical and anesthetic techniques that minimize blood loss with the use of autotransfusions, it is now possible to complete a routine posterior spinal fusion without using allogeneic blood transfusions. The surgeon's role in controlling blood loss begins with selection of the appropriate operation for a given patient.

Preoperative bending films should be used to plan the extent of spinal fusion required. Because blood loss during spinal fusion generally increases as operative time increases, the time necessary to properly install a spinal instrumentation system must be weighed against its advantages.

Operative control of blood loss can begin before the skin incision. When the patient is lying prone the intraabdominal pressure is increased, resulting in vena caval compression which increases the pressure in the venous channels around the spine (Batson's plexus). This can cause increased bleeding during exposure of the spine. Positioning the patient on a device that allows the abdomen to hang free, such as the Relton-Hall frame, avoids the increase in intraabdominal pressure that leads to compression of the vena cava.

Infiltration of the skin and paraspinous muscles with a dilute (1:500,000) solution of epinephrine has been recommended. This is thought to reduce bleeding by producing local vasoconstriction and is claimed to facilitate dissection in the proper plane. There is a transient, although apparently hemodynamically insignificant, rise in serum levels of epinephrine with injection of a 1:500,000 solution. Sponges soaked in epinephrine solution can be packed into the wound to control local bleeding. Careful surgical technique can reduce blood loss. Subperiosteal dissection avoids tearing the numerous vessels in the paraspinous muscles.

Bleeding from bone can be difficult to control but can be minimized by the use of topical hemostatic agents. Bone wax has commonly been used. Although concern has been expressed about bone wax being a foreign material, pathological studies have shown that it is resorbed and replaced by fibrous tissue. Other topical agents such as Avitene (microfibrillar collagen) and thrombin solution have been suggested. The iliac crest, the common donor site for autogeneic bone graft, can be a major source of bone bleeding. Bone wax can be used here as well. Use of cadaver bone graft avoids bleeding from the bone graft donor site entirely. Availability is not a problem, as there are several commercial sources of freeze-dried allogeneic cancellous bone, and many institutions also maintain their own bone banks. Reports suggest that fusion rates are comparable to those with autogenous bone graft.

The primary area of bone bleeding during scoliosis surgery is the spine itself. Excision of the facet joints and thorough decortication of the laminae add to intraoperative blood loss, but increase the likelihood of a successful arthrodesis, the primary goal of the procedure. Thrombin-soaked gelfoam can be temporarily packed into the excised facet joints to aid in hemostasis and removed when the cancellous plugs are inserted. Decortication should be performed as late in the procedure as possible to minimize the time of uncontrollable bone bleeding and allow visualization in the wound for as long as possible. An operation of the magnitude of a spinal fusion will inevitably result in some loss of blood. In patients who cannot or will not receive blood transfusions, such as Jehovah's Witnesses or other situations, it may be necessary to stage the procedure by exposing and fusing a limited part of the spine, stopping when the blood loss reaches a level compatible with normal function without transfusion, and returning to surgery when the patient's blood count has increased.

If a patient is going to receive a blood transfusion, the safest blood is the patient's own: autotransfused blood. Two common methods of autotransfusion are preoperative collection and preservation of a patient's blood, and intraoperative salvage of blood lost in the wound.

Preoperative collection of blood for autotransfusion requires that the patient donate blood to be preserved and refrigerated just like allogeneic donated blood. Patients are allowed to donate one unit per week until 10 days before surgery. Because preservative, citrate phosphate dextrose with adenine (CPDA-1), allows a shelf life as long as 35 days, typically three units can be obtained in the month before surgery. Patients are put on supplemental iron to keep the hematocrit above 30% to allow these donations. Children or patients weighing less than 40 Kg can have proportionately smaller aliquots of blood withdrawn. No special equipment is needed to collect autogeneic blood beyond the resources of a blood bank. It is now the policy to allow parents and other family members to donate blood designated for a certain patient even though such blood is no safer than blood obtained from the general donor pool.

Intraoperative recovery of blood is performed by collection of the drainage from the suction bottles. The red cells are filtered, washed, and concentrated, and then reinfused into the patient. The Cell-Saver system is the most widely used of the several commercially available machines designed for this purpose. The cost of the special equipment and of the technician necessary to operate it has perhaps limited its use.

Both autotransfusion methods are limited in the amount of blood loss that can be replaced. Intraoperative blood recovery systems can recover about 50% of the lost red cells. Donations for autotransfusion are limited by the effectiveness of the preservative, although in special circumstances it is possible to freeze blood for as long as 3 years. If too much blood is lost, allogeneic transfusions are still needed.

Although good surgical technique can minimize blood loss, even the most careful surgeon may find a wide variation in blood loss during what appears to be a standard procedure. To learn more about the mechanisms of blood loss during scoliosis surgery, a study was designed and carried out. All the patients had the same diagnosis (adolescent idiopathic scoliosis), the same anesthesia (hypotensive), the same operation (posterior spinal fusion), and the same surgeon, who used a consistent technique. The working hypothesis was that variations in the patient's hemodynamic status, even under hypotensive conditions, could account for variations in blood loss. The patient's hemodynamic status was monitored with Swan-Ganz, central venous, and radial artery catheters. Because the autonomic nervous system and the renin-angiotensine system both play a major role in hemodynamic regulation, their roles were examined by assaying blood samples for levels of epinephrine, norepinephrine, renin, and angiotensin II. All these variables were measured at set intervals during surgery and whenever the surgeon judged the bleeding to be excessive. Two important findings came from this study: (1) when the surgeon thought there was more bleeding, he was correct; one or more of the hemodynamic parameters had changed; and (2) blood loss correlated most closely with left ventricular stroke work index (LVSWI), a measure of blood flow. Since LVSWI is computed from systemic vascular resistance, cardiac output, and heart rate, an increase in cardiac output at a constant blood pressure, even at hypotensive levels, would increase LVSWI and increase blood loss. With induced hypotension, increases in cardiac output were mediated by increased activity in either the autonomic nervous system, the renin-angiotensin system, or both.

In a broad sense, the surgeon's control of blood loss is related to his skill in controlling the number and size of bleeding points and how long they stay open. Because bleeding from bone is unavoidable, the surgeon's ability to control blood loss is limited. The only thing the surgeon can do is to go faster. The anesthesiologist controls blood flow, or how fast blood is lost through the open vessels. Given the inevitable bone bleeding and a limit to how quickly any surgeon can operate, this means the anesthesiologist becomes the primary determinant of blood loss. The anesthesiologist thus plays an important part in controlling blood loss during spinal surgery. Two major techniques used by anesthesiologists to control blood loss are hemodilution and induced hypotension.

Hemodilution decreases the loss of red cell mass by reducing the hematocrit of the blood lost intraoperatively. The patient is phlebotomized in the operating room and several units of blood are removed and preserved. Circulating volume is maintained by crystalloid replacement. The operation is performed at normal or reduced blood pressure. At the end of the procedure, the patient is diuresed of excess fluid, and the patient's own blood is retransfused. Despite the lower oxygen-carrying capability of the blood, the decreased viscosity of the diluted blood allows better tissue perfusion so tissue oxygenation is maintained. Good results have been reported with hemodilution. It requires no specialized equipment and is simple to perform. However, if one of the goals of minimizing blood loss is to keep the operative field dry as well as to minimize the need for transfusion, intraoperative hemodilution appears less attractive. Blood with a hematocrit of 20 is as difficult to see through as blood with a hematocrit of 40.

Blood loss can also be reduced by decreasing blood flow. Although commonly referred to as hypotensive anesthesia, these techniques are actually directed at reducing the left ventricular stroke work index. Arterial blood pressure is a relatively convenient way of assessing the patient's hemodynamic status, although as previously noted, it is not necessarily the best overall measure of hemodynamic status.

A major concern about hypotensive anesthesia during scoliosis surgery has been the potential for increasing the risk of spinal cord injury. It was feared that hypotensive anesthetic techniques would leave a narrower margin of safety because of reduced spinal cord blood flow. With reduced spinal cord blood flow, an insult to the spinal cord such as distraction or invasion of the spinal canal could be more likely to cause a neurological deficit. However, a growing body of evidence suggests there is little if any increased risk of neurological injury with hypotensive anesthetic techniques. Several investigators have demonstrated experimentally that spinal cord blood flow can be autoregulated independently of systemic blood flow under controlled hypotension.

Somatosensory evoked potentials (SSEP), a noninvasive means of intraoperatively assessing spinal cord function, have been shown not to change significantly with moderate hypotension.

Experimental as well as clinical reports have found that distraction of the spine under moderate hypotension appears no more hazardous than distraction of the spine at normal blood pressure. The anesthesiologist has several alternatives to produce hypotensive anesthesia. High levels of halothane can produce hypotension and reduce blood flow by acting as a vasodilator, by decreasing systemic vascular resistance, and by a direct inhibitory effect on the myocardium, decreasing cardiac output. Like most inhalation agents, halothane anesthesia makes it impossible to monitor SSEPs intraoperatively and to safely and reliably perform a wake-up test. This makes halothane less attractive for routine use in scoliosis surgery. To allow intraoperative spinal cord monitoring and easy reversibility for a wake-up test, balanced anesthesia with nitrous oxide supplemented with narcotics has become popular. Reduction of blood flow can be obtained with intravenously administered hypotensive agents. The two agents most commonly used are sodium nitroprusside and trimethaphan.

Sodium nitroprusside is a peripheral vasodilator that relaxes the smooth muscle in the walls of the arterioles and venules. It is administered by constant infusion. Sodium nitroprusside-induced hypotension is difficult for the anesthesiologist to regulate. The blood pressure varies, and the infusion rate must be constantly altered. The induced hypotension quickly reverses when the infusion is stopped. Sodium nitroprusside does not block the catecholamine release that occurs in response to diminished blood pressure and which can cause rebound hypertension with increased bleeding. Rebound hypertension may be noted intraoperatively, but more typically will be evidenced in the recovery room by excessive drain output and a blood-soaked dressing. While the rebound effect can be partially blocked with beta blockers such a propanalol, there is little advantage to the patient in decreasing intraoperative blood loss only to increase blood loss that night or the next day.

Trimethaphan acts as both a peripheral vasodilator and a ganglionic blocker, so rebound hypertension is not a problem. Trimethaphan is also administered by constant intravenous infusion. Unlike sodium nitroprusside, trimethaphan-induced hypotension does not reverse quickly when the infusion is stopped.

While both sodium nitroprusside and trimethaphan have been used clinically with good results, there are some experimental differences between these hypotensive agents and their effect on spinal cord blood flow. Spinal cord blood flow in dogs was decreased for the first 30 minutes of sodium nitroprusside-induced hypotension. As autoregulatory mechanisms came into play, spinal cord blood flow then returned to normal levels. With sodium nitroprusside-induced hypotension, there were no deleterious effects from spinal distraction once the spinal cord blood flow had returned to normal. In contrast, spinal cord blood flow during trimethaphan-induced hypotension decreased and paralleled the decrease in mean arterial pressure. There was no autoregulatory compensation observed with trimethaphan. When the trimethaphan infusion was stopped, there was a slow return of spinal cord blood flow to normal. However, distraction of the spine did not cause any problems even with this diminished spinal cord blood flow.
Experiments with nitroglycerine-induced hypotension in dogs have shown that spinal cord blood flow was maintained at control levels throughout the period of induced hypotension. This is potentially a great advantage. It is uncertain how much data from the canine spinal cord blood flow model can be extrapolated to clinical cases. However, regulatory mechanisms of blood flow in the spinal cord appear to be similar in most mammals.

Controlling blood loss during scoliosis surgery is a challenge to both the surgeon and the anesthesiologist. Both have other concerns during the procedure. In addition to minimizing blood loss, the surgeon's primary goal is a safe, successful arthrodesis. The anesthesiologist's technique must allow intraoperative assessment of spinal cord function either by SSEP or a wake-up test or both. Successful control of blood loss allows the patient to receive a better operation and significantly reduces the patient's risk of acquiring transfusion-related illnesses.

The surgical technique for posterior spinal fusion is divided into two parts. The first part discusses exposure and arthrodesis of the spine, independent of the type of instrumentation. The second part discusses techniques for implanting different instrumentation for correction of scoliosis.

A straight-line incision gives a better cosmetic result. Skin scratches should be avoided. A heavy suture pulled taut from the spinous process of T1 to the gluteal cleft can be pressed lightly to the spine and the planned incision marked with a skin scribe. The skin and subcutaneous tissues can be infiltrated with a 1:500,000 solution of epinephrine to facilitate hemostasis. The incision is carried down to the fascia overlying the spine. Bleeding can be controlled with electrocautery and bipolar coagulation. Self-retaining retractors should be used to keep the tissues under tension and facilitate dissection in the proper plane. Superficial dissection continues until the tips of the spinous processes are palpable in the slight depression between the paraspinal muscles. It may be necessary to undermine the skin slightly to palpate the spinous processes at the apex of the curve. Using a Kelly clamp, each spinous process is straddled while the fascia over it is divided with coagulating cautery. The fascia between the spinous processes is divided by "connecting the dots." Subsequent dissection takes advantage of the thicker periosteum in children and adolescents. A careful subperiosteal dissection reduces blood loss and can usually be accomplished without difficulty. Dissection should start at the caudal end of the wound and proceed cranially to facilitate stripping muscle fibers from their origins on the spine. Using a Cobb elevator, the surgeon should start on the side of the spinous process, come down onto the lamina, and sweep toward the facet joints. Sponges soaked in 1:500,000 epinephrine solution are packed in the wound to facilitate hemostasis. If there is bleeding from the bone, electrocautery or bone wax is used to control it. The assistant alternates dissecting with the surgeon, stripping and packing on one side of the spine, then on the other. At that point a check radiograph should be obtained to determine the levels exposed. A towel clip is placed on the base of a spinous process. The spinous processes in the thoracic spine point caudally, and a towel clip placed near the tip of the spinous process may be superimposed over the next caudal vertebral body on the radiographs. Placing the towel clip near the thoracolumbar junction is helpful since the ribs are convenient landmarks.
No attempt should be made to expose the tips of the transverse processes on the first pass or excessive bleeding will result. Once the entire extent of the spine has been partially exposed, the epinephrine-soaked sponges are removed, self-retaining retractors are replaced deeper in the wound, and a second pass is made beginning caudally and proceeding cranially. On the second pass, soft tissue should be stripped from the spine to the tips of the transverse processes. The interspinous ligaments and facet joint capsules in the lumbar spine are removed with a combination of rongours and curettes, taking care to preserve the facet joint capsules and interspinous ligaments between the last vertebra in the fusion and its caudal neighbor. Working up into the thoracic spine, curettes are used to work along the superior and inferior aspect of each lamina to expose the facet joint.

Meticulous hemostasis is essential. Bleeding should not be excessive if the dissection remains subperiosteal. If there is excessive bleeding, the anesthesiologist should be informed because the blood pressure and cardiac output may have increased. Now that the entire spine over the levels to be fused has been subperiosteally exposed to the tips of the transverse processes, all the facet joints are visible and the soft tissues are held back by self-retaining retractors.

The next step is ablation of the facet joints. Care should be taken to preserve, for the time being, the facet joints at sites where hooks are to be inserted. There are two popular techniques for ablation of the thoracic facet joints. Moe described directing a gouge onto the facet obliquely from medial to lateral, raising a flap off the inferior aspect of the lamina to expose the superior facet joint of the caudal vertebra. The facet can then be denuded of cartilage with a gouge or curette and a chunk of cancellous bone can be lightly impacted into the joint. Hall described using a Capener gouge directed straight down on the facet joint on the inferolateral aspect of the lamina, exposing the superior facet joint of the vertebra below. The articular cartilage can be curetted or gouged out. Again, a chunk of cancellous bone should be inserted. The lumbar facets can be excised with a carefully directed osteotome and curette. Instead of placing bone graft in each facet joint as it is excised, thrombin soaked gelfoam can be packed in to aid in hemostasis. It should be removed before bone graft insertion.

The ideal bone to use for posterior spinal fusion is autogenous bone. Most commonly, bone graft is harvested from the posterior iliac crest. If the fusion extends to the lumbar region, the posterior iliac crest can be exposed by caudal extension of the incision and dissection above the lumbar dorsal fascia. A separate incision along the posterior iliac spine can be made if a thoracic level fusion is being performed. The posterior iliac crest is split sharply and the outer table is then exposed subperiosteally. The outer table is removed and strips of cancellous bone taken using a gouge. Care should be taken to avoid penetration of the sacroiliac joint near the posterior inferior spine of the iliac crest. The superior gluteal artery passing through the sciatic notch should be avoided. Once the bone graft is harvested, bone wax can be laid over the raw surfaces to aid hemostasis. A drain is inserted; a large towel can be packed in the wound, and the fascia over the crest can then be towel clipped shut until closure.

To this point, the exposure of the spine has been made in a standard fashion regardless of instrumentation.

Both the Harrington and Cotrel-Dubousset (C-D) systems require distraction hooks be placed on the inferior aspect of upper thoracic lamina and the superior aspect of lower thoracic or lumbar lamina. Both may also require compression hooks to be placed around transverse processes in the upper thoracic spine or on the inferior aspect of the lamina in lower thoracic or lumbar vertebrae.

The C-D system has special instruments to facilitate exposure of the hook sites, but the principles remain the same. To prepare a distraction hook site in the thoracic spine, a narrow osteotome is used to make a cut in the lamina parallel to the long axis of the spinous process. The cut should carefully go through both cortices. A second cut is made perpendicularly removing the inferolateral corner of the lamina. This should expose the underlying superior facet of the next caudal vertebra. The osteotome used to make these cuts should be very sharp and should be controlled carefully to avoid penetration into the spinal canal. Some surgeons direct the second cut slightly cranially toward the midline to reduce the risk of lateral hook dislodgement. The space under the lamina can then be opened with a small Penfield-style elevator, a sharp Harrington hook, or the pedicle finder from the C-D instrumentation set. Care should be taken to get below both cortices of the lamina. The aim is to release any remaining soft tissue and to locate the pedicle if desired for hook placement. The intended hook can be inserted and gently impacted into place.

For placement of compression hooks in the upper thoracic spine, an elevator or a sharpened hook is used to strip the soft tissue from the transverse processes. In the lower thoracic spine the compression hooks must be placed sublaminarly because the transverse processes are usually too small to hold the hook. The laminotomy is started with a Leksell rongeur, biting away the ligamentum flavum in the midline until epidural fat is visible. Part of the downward-pointing spinous process from the superior vertebra usually needs removal as well. Any bone removed can be saved for grafting. A Kerrison rongeur is used to remove the rest of the ligamentum flavum and square off the inferior aspect of the lamina for better seating of the hook.

In the lower thoracic lumbar spine, laminotomies are needed for insertion of compression or distraction hooks. The ligamentum flavum is excised, exposing the edge of the lamina. A laminar spreader may facilitate exposure. A Kerrison rongeur is used to make a square cut in the lamina. If instrumentation is to be carried out in the lower lumbar spine it is often helpful to use a large Leksell rongeur to thin out the lamina immediately above the hook site. This will facilitate pushing the rod low enough to seat the hook completely below the lamina. If a lamina is damaged during the preparation of a hook site, an adjacent lamina, superiorly or inferiorly, can be prepared for hook placement.

Installation of the two hooks for the ratcheted distraction rod requires laminotomies with minimum invasion of the spinal canal. Some surgeons use the forked hook designed by Zielke to straddle the pedicle, which minimizes penetration into the spinal canal. Once both hooks have been placed, the clothespin-like Zielke hook clamps can be attached to the hooks. The Harrington distraction outrigger can then be attached to the Zielke hook clamps. Using the outrigger, the spine can be repeatedly distracted to overcome the viscoelasticity of the spine and gradually gain more correction over a period of time. With the outrigger in place, excision of the facet joints can be performed, often resulting in further mobilization of the spine and additional correction. While the outrigger is under tension, the wound can be packed with sponges and the iliac crest bone graft can be harvested. By the time the bone graft has been harvested, the outrigger can usually be further distracted.

If a square-ended Harrington rod is to be used, it should be contoured appropriately to preserve lumbar lordosis and thoracic kyphosis. With the Zielke hook holders holding the outrigger, it may be possible to insert the Harrington distraction rod with the spine distracted. If the rod is contoured a great deal, rod insertion with the distractor in place may be difficult. Removal of the distractor, expeditious insertion of the rod, and redistraction may be less frustrating and time consuming. Care should be taken using the Harrington rod spreader to apply force gradually to minimize the risk of lamina fracture. Once the distraction rod has been inserted, the compression apparatus can be inserted if desired. The Keene modification of the compression system allows placement of all needed bushings and nuts on the threaded, flexible compression rod. The hooks are placed under the laminae or transverse processes and the rod dropped into the top opening slot. A bushing is slid along the rod to engage each hook, locking the rod down.

After all the bushings are engaged, the hexagonal nuts are tightened. Care should be taken when tightening the compression apparatus because the wrench does not give the sensory feedback that the distractor gives; the first sign of overtightening may be fracture of the transverse process. In the thoracic spine, overtightening can reduce thoracic kyphosis. After the compression rod is inserted, it can be cross-linked to the distraction rod by the transverse loading device of Cotrel or wire loops for additional stability. After installation of Harrington rods, a generous area of lamina remains accessible for decortication of bone grafting.

After the Harrington distraction hooks and outrigger are placed, angled awls are used to drill a hole in the base of each spinous process. A doubled Wisconsin wire is passed from each side and threaded through the button of the opposite wire. Each wire is pulled taut against the base of the spinous process and the button can be lightly impacted into place with a punch. The wire loops are opened up on the concave side, and a previously contoured Harrington rod is inserted. The spine is distracted and the wires twisted loosely to approximate the rod to the spine. Next, a contoured Luque rod with at least one L-bend is passed through the open loops on the convex side of the curve. The wires are loosely twisted around the Luque rod. The distraction rod is retightened. Each rod is then pushed to the spine and the wires tightened.

Use of C-D instrumentation is very challenging, and supervised, hands-on training is essential. The components give the spinal surgeon tremendous flexibility and allow use of the C-D for virtually any condition or problem in the spine. For scoliosis surgeons, the major appeal of the C-D system is its purported ability to correct thoracic scoliosis while maintaining or restoring thoracic kyphosis and improving rotational alignment of the spine. The technique for C-D instrumentation of a thoracic scoliosis is presented here. The spine is exposed in standard fashion. Hook sites on both sides are prepared. The facets are excised and grafted and the laminae decorticated on each side before the rod is inserted. Routine C-D instrumentation of thoracic scoliosis usually requires placement of eight hooks and two rods. Correct hook placement is essential. Selection of hook sites is based on the preoperative anteroposterior, lateral, and side-bending radiographs of the spine. The neutral vertebrae are at the ends of the curve and they should be in neutral rotation; that is, the spinous process should be halfway between the pedicles. The apical vertebra appears most rotated and has a vertical convex border. The intermediate vertebrae are the end vertebrae of the rigid portion of the curve, which usually extends over four or five levels with the apical vertebra in the center. The concave side is instrumented first. Hook sites are prepared on the two neutral and two intermediate vertebrae. Hook selection is as follows: for the cephalad neutral vertebra, a closed pedicle hook; for the cephalad intermediate vertebra, an open pedicle hook; for the caudad intermediate vertebra, an open laminar hook; and for the caudad neutral vertebra, a closed laminar hook. Depending on the size of the patient, the laminar hooks can be the larger lumbar or smaller thoracic hook. The cranial two hooks point cranially; the caudal hooks point caudally because this rod is intended to distract the spine. Using a template, if necessary, a rod of appropriate length is contoured to roughly parallel the scoliosis. As the rod is rotated, it will in a sense convert the thoracic scoliosis to thoracic kyphosis. Excessive contour in the rod makes it easier to seat it on the hooks, but makes it more difficult to spin the rod and restore thoracic kyphosis. Insufficient rod contouring will reduce the possible correction of rotation and will leave the rod functioning more like a Harrington distraction rod. Inserting the rod and sliding the blockers into place is similar in principle to applying the Wisconsin compression apparatus with its open hooks and bushings. However, the C-D rod is not flexible, and the hook blockers must be rotated as well as pushed along the rod to seat in the open hooks. Fortunately the C-D instrument set includes special instruments for facilitating rod placement in the hooks. These include rod and blocker pushers as well as a hook clamp with a lever that can pull the seated hook up to the rod. Note that fracture of the lamina can occur if excessive force is used.

After the rod and the hook blockers are inserted, C-rings are placed behind the blockers at the two intermediate vertebrae. The C-rings should be placed so that they can easily be removed when they are spun 90 degrees. The lock nut on one C-ring is tightened; then the rigid segment is distracted by pushing against the other C-ring, which is then locked into place. The hooks in the neutral vertebrae at the ends of the curve are left loose to allow the rod to spin inside them. Using the two large rod clamps, the rod is slowly rotated while the C-rings keep the intermediate hooks in place. As the contoured rod rotates, the spine is straightened in the frontal plane, the thoracic scoliosis is reduced, while the spine is contoured more in the sagittal plane, the thoracic kyphosis is restored.

The flexibility and degree of curvature of the spine and the contour of the rod affect how much correction can be obtained. As with distraction of a Harrington rod, gradual correction to overcome the spine's viscoelasticity is important. Excessive rotation may result in lamina fracture and pulling out of the intermediate hooks. When the desired rotation has been achieved, the lock nuts on the two hook blockers are tightened and the C-rings removed. Additional distraction is applied to the two neutral vertebrae hooks to further correct the scoliosis. The spine may now look almost straight. The nuts on all the hooks and blockers are now tightened.

Placement of the convex rod is designed to push the lamina at the apex of the curve more anteriorly, further derotating the spine while reducing the height of the rib prominence. Hook placement for the convex rod is as follows: on the cranial neutral vertebra, a "claw" consisting of a closed laminar hook around the transverse process and a closed pedicle hook; on the apical vertebra, an open pedicle hook; and on the caudal neutral vertebra, a closed laminar hook. Again the choice of lumbar (large) on thoracic (small) laminar hooks depends on patient size. The convex rod is contoured with bends about one-fifth of the way from each end. The upper fifth is bent into kyphosis, the lower fifth into lordosis. These bends result in a rod with a gentle "S" curve when viewed from the side. The rod and blocker for the open pedicle hook is seated. The two hooks of the claw are pulled together and locked to the rod. Compression is applied at the apical and distal hooks, resulting in further correction of the scoliosis.

If desired the wake-up test can now be performed.

The nuts on all the hooks and blockers are tightened until they shear off, locking everything in place. The two rods are linked with the device for transverse traction (DTT) proximally and distally to increase the rigidity of the construct. Instrumentation of double curves requires more hooks, using each rod to distract the concave side of a curve and compress the convex side. To ensure a solid arthrodesis of the spine, decortication precedes rod placement. With all the hooks and blockers, rod placement, rod rotation, and tightening the hooks can be time consuming, possibly resulting in increased blood loss with C-D instrumentation.

After the spine has been corrected to its maximum extent, a wake-up test, or clonus test, can be performed. The level of anesthesia is lightened while the operative field is flooded with warm saline to minimize risk of air embolism. Once the wake-up test is successfully completed, anesthesia can be reinstituted and wound closure can begin.

The muscle and fascia are closed with interrupted stitches of absorbable suture. Drains below the fascia, directly against the bleeding bone, will not allow formation of a hematoma that can help tamponade further bleeding. Also, a drain is placed above the fascia and brought out through a stab wound. To leave a better cosmetic result the skin is reapproximated and closed with a running subcuticular stitch.

The most dreaded intraoperative complication of spinal surgery is paralysis. MacEwen reported a 1:400 incidence of spinal cord injury with scoliosis surgery. The risk of neurological injury may be increased with excessive distraction and with certain devices, such as sublaminar wires. Injury to the spinal cord from overdistraction is thought to be on the basis of compromise of spinal cord blood flow. Injuries from invasion of the spinal canal are probably from direct trauma.

Increased latency and reduced amplitude of the SSEP signal may indicate neurological compromise, but may also be the result of excessive spinal cord cooling or inappropriate levels of inhaled anesthetic agents; irrigating with warm saline or altering the anesthesia may return the signal to normal. Alteration of the SSEP signal or failure of a wake-up or clonus test require immediate action. Reduction of distraction by loosening or removal of the rod should be quickly performed. It may be advisable to discontinue hypotensive anesthesia to allow maximum blood flow to the spinal cord. The role of steroids in possible injuries to the spinal cord remains controversial, but it is preferable to use it before the operation.

When too large a force is applied, the lamina or transverse process may fracture, requiring an adjacent vertebra to be instrumented. Lamina fracture is probably more common in the thoracic spine where smaller laminae and smaller hooks result in a larger concentration of forces. If a wire is overly tightened, breakage can result. Breakage of Wisconsin wire is usually not a problem. The wire can often be replaced, and if not, one less wire does not decrease stability much. If a sublaminar wire breaks, acute removal is controversial. Leaving a broken sublaminar wire in place may allow the wire to spring back into the canal somewhat like opening up a safety pin. But the wire may also spring more deeply into the canal as it is removed. We recommend removal of the wire if possible. Improper contouring of a rod may not become apparent until after insertion. Special instruments are available to help recontour the rod in situ. Alternately, the rod can be removed and a new properly contoured rod inserted.

Excessive bleeding can be the result of poor surgical technique but may be a consequence of poor anesthetic control of blood flow. Any sudden increase in blood loss without obvious cause should be brought to the attention of the anesthesiologist. If excessive bleeding continues, it may be judicious to terminate the procedure, inserting any available bone graft. Intraoperative cardiac arrest, which has been rarely reported, may result from hypovolemia, air embolism, or transfusion reaction. Air embolism is thought to result from an increased negative intrathoracic pressure that can pull air in through the exposed sinusoids of the laminae and result in a large bolus of air going to the heart. The wake-up test is a period of increased risk because the patient's ventilation is less controlled and the patient may breathe spontaneously. Flooding the field with saline before the wake-up test helps prevent this catastrophic complication.

A dural tear can result in cerebrospinal fluid (CSF) leak. If possible, the tear should be immediately repaired. If the tear is inaccessible it is often sealed off by the hematoma from the wound. If a dural tear occurs the patient should be kept flat in bed for 5 days postoperatively. This prevents the rise in CSF pressure that accompanies the upright position, which could reopen the tear.

While still intubated, the patient is carefully rolled onto a bed or stretcher. The patient is extubated in the operating room or recovery room. The instruments should be kept sterile, and the operating team should be prepared to reopen the wound and remove the instrumentation in the event the patient demonstrates loss of neurological function on awakening. If the fusion included the thoracic spine, a portable radiograph is taken in the recovery room to rule out a pneumothorax.

Postoperative use of the Stryker frame has decreased as the stability of internal fixation of the spine has increased. Patients are log rolled and positioned with pillows on a regular bed. Patients are kept on their backs for about 6 hours after surgery in an effort to reduce bleeding by maintaining pressure on the wound.

Serial neurological examinations should be performed in the first 24 hours postoperatively because not all neurological complications become apparent immediately post-operatively. Epidural hematoma may result in spinal cord compression and delayed paraplegia. Postoperative facial edema is common from the dependent position and will resolve as the patient mobilizes his/her third-space fluid. Improper positioning of the patient can result in postoperative nerve compression syndromes. The lateral femoral cutaneous nerve of the thigh can be compressed from the Relton-Hall frame, resulting in lateral thigh numbness. The brachial plexus may be stretched if the arms are abducted or extended excessively during surgery.

Many patients who undergo posterior spinal fusion for idiopathic scoliosis develop the syndrome of inappropriate antidiuretic hormone (SIADH) in the immediate postoperative period. The exact etiology is poorly understood, but the increase in antidiuretic hormone results in a decrease in urine output despite decreasing serum sodium levels. The danger is that SIADH may be mistaken for a hypovolemia and excessive intravenous fluid may thus be administered, resulting in fluid overload and pulmonary edema. SIADH tends to present in the middle of the night when the operating surgeon may not be readily available and the patient may be under the care of a junior house officer. SIADH usually resolves without any treatment and fluid restoration is seldom necessary. If urine output is unsatisfactory, a small amount (5 mg) of lazix will restore good urine flow.

Superior mesenteric artery (SMA) syndrome is occlusion of the third part of the duodenum by the superior mesenteric artery at its oblique takeoff from the aorta. The lengthening of the trunk seen with distraction instrumentation is thought to be a contributory factor. Thin patients seem more susceptible, perhaps because they have less retroperitoneal fat to cushion the duodenum. Affected patients present with symptoms of an upper gastrointestinal tract obstruction such as vomiting. Treatment consists of nasogastric suction until the obstruction resolves, usually in a few days. Occasionally hyperalimentation is needed if the obstruction persists.

Postoperative wound infections are uncommon in scoliosis surgery. The preferred treatment for a deep wound infection is thorough debridement and closure over drains. Removal of the hardware is not always necessary and may be contraindicated. Fusion can proceed in spite of infection, and is probably more likely to occur if motion of the spine is kept to a minimum.

The need for postoperative immobilization depends on the type of internal fixation, how well it is inserted, and the preference of the surgeon. Some forms of internal fixation are claimed to be stable enough that no external immobilization is necessary. It is preferable to keep the  patients be fitted with a plastic bivalved thoracolumbosacral orthosis (TLSO). After the TLSO is fitted, the patient is mobilized, usually by about the fourth or fifth day postoperatively, and discharged from the hospital about 7 days postoperatively. By this time the patient is able to walk and climb stairs. The patient is instructed to wear the brace at all times except when taking a shower and is restricted to lifting not more than 5 Kg. Most patients are able to return to school or work by 3 to 4 weeks after surgery. The brace is worn for 6 months postoperatively. Activity is restricted for 1 year, at which time resumption of normal activities is permitted with the exception of gymnastics and contact sports such as football, hockey, and wrestling. Follow-up radiographs are taken at 1, 3, 6, and 12 months, and then yearly for 5 years.

Relaxation in the tension in the soft tissues of the spine results in a gradual loosening of the instrumentation. Typically this results in only a minimum (less than 5°) loss of correction, and usually occurs in the first few months after surgery. Loosening or dislodgement of an internal fixation device can result in a substantial loss of correction. Breakage of an internal fixation device is usually a fatigue failure, and strongly suggests the presence of pseudoarthrosis. The pseudoarthrosis may not be readily visible on routine radiographs, and tomograms may be necessary to locate the exact site. The need for reoperation depends on the patient's symptoms, the degree of deformity, and any progression in deformity. Revision of the fusion with additional bone grafting is usually successful.

The use of an insufficiently contoured rod into the lower lumbar spine can result in loss of lumbar lordosis and the so-called flat back syndrome. Affected patients walk with an awkward, crouched gait. In fusions extending into the lower lumbar spine, great care should be taken to preserve lumbar lordosis by contouring the rod. Posterior translation (retrolisthesis) of the vertebra at the caudal end of a distraction rod has been reported to be associated with an increased incidence of pain in one long-term follow-up study. To minimize the occurrence of retrolisthesis, Bradford described passing a wire loop around the two most caudad spinous processes before distraction. A spinal fusion puts an increased load on the disc and at the end of the fusion. This increased load can lead to premature disc degeneration, degenerative facet hypertrophy, and spinal stenosis. At present these problems are mostly associated with fusion to the lower lumbar spine. However, the modern era of spinal fusion dates back only 25 years. The adolescent girls who underwent fusions ; then are now only entering their fifties. It is possible .¬that with more time there will be more problems.

The operative management of idiopathic scoliosis has been discussed in detail. Other forms of scoliosis may require different techniques.

Operative management of congenital scoliosis is in many ways similar to management of idiopathic scoliosis. Instrumentation to correct the deformity is used less frequently, because there appears to be an increased association with incidence of neurological complications. If the use of corrective instrumentation is contemplated, a preoperative MRI is advisable to rule out problems such as diastematomyelia. If the goal of surgery is prevention of further progression, an in situ fusion may be sufficient. Special care is needed to determine intraoperatively the levels to be fused because the anatomy is distorted. Usually less of the spine is fused in cases of congenital scoliosis; often the abnormal vertebra(e) is fused to the two adjacent normal vertebrae. Compensatory curves should not progress once the congenital curve is fused. After the segments to be fused have been exposed and decorticated and the facet joints excised, metallic clips are placed on the spinous processes of the cranial and most caudal vertebrae to mark the extent of fusion and to serve as references for measuring growth and possible curve progression. Autogenous bone graft should be used whenever possible. Postoperative immobilization with either a cast or brace is maintained for 6 to 12 months.

Scoliosis in neurologically impaired patients presents special challenges to the spinal surgeon. Patient cooperation is often unobtainable. The bone in the spine may be of poor quality because of disuse osteoporosis, the use of antiseizure medication, poor nutrition, or a combination of all these. Postoperative immobilization is often difficult to maintain. Patients with neuromuscular diseases often have poor temperature-regulating mechanisms. The use of heated, humidified anesthetic gases can help prevent excessive heat loss during surgery. The need for secure fixation of the spine with minimum postoperative immobilization is often best served by the use of segmental spinal instrumentation, that is, an instrumentation system with multiple attachments to the spine. The three major forms of segmental spinal instrumentation are (1) the sublaminar wires and L-rods popularized by Luque; (2) the intraspinous process wires of the Wisconsin system; and (3) the use of multiple hooks of the Cotrel-Dubousset (C-D) system. (Use of the C-D system and Wisconsin wires has already been discussed in the section on idiopathic scoliosis. )

Luque or L-rod instrumentation consists of two smooth, contoured, L-shaped rods attached to the spine with multiple sublaminar wires. Rods of :3/8-in. and 1/4-in. diameter are available. The smaller rods are easier to bend and cut, but the larger rods are twice as strong. Luque recommended contouring the rods preoperatively based on bending films, but contouring can also be done intraoperatively. It is important to realize that rarely can perfectly straight rods be inserted without increasing the risk of neurological injury from overcorrection of the deformity. An exception to this is the collapsing curve seen in muscular dystrophy, which can be well managed with Moseley's unit rod or two straight L-rods. In addition, the rods should be contoured in the sagittal plane to preserve thoracic kyphosis and lumbar lordosis.

The exposure of the spine has been described. The spine is exposed subperiosteally to the tips of the transverse processes; the facet joints are excised and bone grafted. Good hemostasis is essential to minimize the risks of sublaminar wire passage. Passage of the sublaminar wires requires removal of the ligamentum flavum at each level and usually a small laminotomy as well. In the thoracic spine the downward-pointing spinous processes often need to be removed. A transversely directed large rongeur can be used to carefully bite down in the interspace between two laminae and remove the superficial layers of ligamentum flavum just until the epidural fat shows through. Then a smaller Leksell or Kerrison rongeur can remove the remainder of the ligamentum flavum. Bleeding epidural vessels can be coagulated with bipolar cautery or packed off with thrombin-soaked gelfoam. A doubled 16- or I8-gauge wire is contoured into an arc just long enough to span the lamina with a rounded, slightly upwardly bent tip. Wires are passed most easily in a caudal-to-cranial direction with a continuous gently upward pull to avoid too deep an incursion into the spinal canal. The tip of the wire should always be in contact with the undersurface of the lamina. If there is resistance, the wire should be withdrawn and a new attempt made. When the tip appears in the next interspace, it should be firmly grasped with a large needle holder or similar instrument. While maintaining a constant upward pull, the wire is carefully advanced until half the length is on each side of the lamina. The tip is cut off (taking care to ensure it does not fall in the wound) and the two loops of wire are pulled apart, one to each side of the spine. From this point on, care must be taken to avoid jostling the wires and pushing the loops further into the spinal canal. To help keep the wires from unintended penetration into the spinal canal, the loop can be crimped over the lamina, then bent back out of the wound and secured with a clamp. Yngve described looping the wire on itself to accomplish the same thing, prevention of spinal cord injury by accidental jostling of the wires.

To insert the contoured L-rod, the two ends of each wire can be pulled to opposite sides of the wound and the rod laid along the laminae. Alternately, the clamped loops of wires can be pulled open and the rod threaded up. While rod pushers push the rod to the spine, the wires are tightened, starting at the apex of the curve. Pulling the rod to the spine by tightening the wires is likely to result in wires cutting through the lamina. A distraction rod can be substituted for the contoured Luque rod. This so-called Tex/Mex or Harri-Luque system allows for axial distraction as well as transverse correction. Once both rods are inserted, the wires are retightened and the two rods are tied together at 7 levels by wire loops.
In neuromuscular scoliosis, the fusion may need to be extended to the pelvis. Luque rods can be fixed to the pelvis in several ways. The "L" of the rod can be pushed through a hole drilled transversely across the posterior ilium, just above the sacrum, similar to implanting a Harrington sacral bar.

More secure pelvic fixation results from the "Galveston technique". The end of the rod is driven into a hole predrilled in the ilium just above the sciatic notch. A series of bends over the sacrum direct the remainder of the rod up the spine. This technique is quite demanding, and preoperative bending of the lower portion of the rod is recommended. Less bone graft is available after installation of L-rods with the Galveston technique, so excision and grafting of the facet joints is important to secure a solid arthrodesis.

The lack of bony elements posteriorly in this condition may require an anterior approach with discectomy and interbody fusion as an initial procedure. Dissection through the scar tissue posteriorly may be difficult. Cases where tethering of the spinal cord must be considered and corrected accordingly.

The paucity of posterior elements in the lower part of the spine in the myelomeningocele patient presents special challenges. Segmental instrumentation is often possible, either by passing Luque wires around the everted laminae or by carefully mobilizing the dura and passing Wisconsin wires through the upturned laminae. Luque rods can be extended into the pelvis by the Galveston technique, or a contoured Harrington rod with a sacral hook can be inserted onto the sacral ala. Many children have a pelvis too small for a Moe-style sacral hook, and a Leatherman hook can be used instead. Above the dysraphic levels of the spine, sublaminar or spinous process wires are used. Lumbar lordosis must be maintained to allow the child to sit forward on his thighs.

Special postoperative complications in myelomeningocele include acute hydrocephalus, particularly if the cord is transected and oversewn. Surgery of this magnitude can result in shunt malfunction for reasons poorly understood, and excessive lethargy postoperatively should be investigated. The rate of infection and pseudoarthrosis are higher in this group of patients; if postoperative immobilization is needed, a removable body jacket allows for inspection of the skin.

Posterior spinal fusion for childhood spinal deformity requires careful preoperative planning, familiarity with the anatomy of the spine, and familiarity with the planned instrumentation. Close cooperation with the anesthesiologists intraoperatively is essential to minimize the need for blood replacement. If these steps are taken, a variety of childhood spinal deformities can be safely managed.