The available sizes of interspinous spacers correspond to measurements of the interspinous spaces published in two studies [59, 60]. These studies suggest that the spacers should be placed anteriorly in the interspinous space, where the cortex is thicker and the interspinous process distance is greatest [60]. An in vivo radiographic study found smaller interspinous spaces, suggesting that bone trimming is typically necessary to avoid creating undesired kyphotic changes in segments treated by interspinous dynamic stabilization [61], confirming our own experience.

Maintaining segmental lordosis is capital and merits a short digression regarding several important operative and postoperative aspects: first, in order to select a spacer size that will not induce kyphosis, the patient must be operated in the prone position, not in the genu pectoral position. To insert the Wallis or the UniWallis when the patient is in the prone position, one may use a distractor. If the device fits tightly when the patient is in the genu pectoral position, the bands will loosen whenever the patient is upright. This will certainly not contribute to restoring segmental stiffness! The spacer should be pushed in against friction, but not be large enough to induce kyphosis. It is also important to use a spacer that is not too small. If it is too small, there is a risk of excessive movement that might scratch the spinous processes. Another operative point concerns the supraspinous ligament, the distal end of which is L4 in the majority of subjects. I always thought it important to disinsert the fascia of the paraspinal muscles from the spinous processes after the initial incision, then, at the end of the Wallis implant procedure, to reattach them with a suture through a small hole at the tip of the spinous process. It is also important to insist that the patient must wear a lumbar corset for 1 month following the operation to optimize development of scar tissue around the implant. Finally, just as one should do in selecting patients for fusion, surgeons should take into account psychological, social, or workers compensation-related issues when deciding which patients would best benefit from dynamic stabilization (end of digression).

In interspinous dynamic stabilization implants that do not have flexion-limiting bands, there is conflicting biomechanical evidence. In one study with axial preloading applied to cadaver specimens of the lumbar spine, Coflex, In-Space, and Aperius reduced range of motion (ROM) of the implanted segment in both extension and flexion even though the supraspinous ligament was removed for placement of the Coflex [62]. In a similar study with preloading, In-Space reduced ROM in extension, but not in flexion, lateral bending, or axial rotation; reduced disk pressure of the operated segment, with little effects on ROM or disk pressure in the adjacent segments [63]; and unloaded the facet joints when loaded in extension [64]. In contrast, cadaver studies of spacers alone (without tension bands) without axial preloading show significant limitation of ROM in extension, but not in flexion, as well as no influence on axial rotation or lateral bending [65–67]. This would apply less if the stabilization devices could be implanted without injury to the thoracolumbar fascia, fascia of the longissimus thoracis muscles, or fascia of the multifidus muscles, although avoiding these injuries is difficult when decompressing the lumbar canal [68]. Regarding the segments adjacent to implantation by a spacer alone, Hartmann et al. reported that flexion-extension ROM above and below was increased in most of their cadaver specimens with and without a follower load of 400 N [69]. Mao et al. reported increased ROM only in extension at L2–L3 with a Coflex at L3–L4, which limited flexion-extension ROM there with little effect on axial rotation or lateral bending, adjacent to L4–L5 with rigid pedicle screw fixation [70]. An in vivo radiographic study has reported that limitation of flexion by the Coflex device is possible in case of bone overgrowth into the device, but generally not expected [71]. Coflex can be used to strongly limit flexion, but the solution calls for rivets through the spinous processes, with high risk of spinous process fracture especially in L5 and in patients with poor bone quality [72].

In an ovine model, Gunzburg et al. showed that an interspinous spacer reduced total segmental ROM in flexion-extension by only 17 % and that the combination of a spacer plus a tension band around the spinous processes reduced the ROM by 46 % [19]. Contrary to some interspinous dynamic stabilization devices, all three generations of our interspinous stabilization devices have a strong cord or band that limits flexion. A recent cadaver study with application of compressive preloads mimicking the stabilizing action of axial musculature demonstrated that anatomical alterations corresponding to degenerative and iatrogenic lesions result in decreased segmental stiffness. This loss of stiffness was less amenable to compensation by axial muscular activity in flexion than in extension, suggesting to the authors the potential usefulness of surgical implants that specifically increase flexion stiffness and limit flexion ROM to counteract the iatrogenic instability resulting from surgical decompression [4]. Without preloading, other cadaver studies have shown that the Wallis system restored ROM to that of the intact specimen in both flexion and extension, while Coflex and DIAM restored ROM only in extension, but not in flexion; all three systems reduced lateral bending by 10 % and axial rotation by 20 % [66, 67]. In a randomized trial of patients operated at L4–L5 for herniated disks or canal stenosis by either PLIF or Wallis dynamic stabilization, Li et al. measured the stiffness of degenerated L4–L5 segment. L4–L5 stiffness before decompression was 37 Nm; after decompression, it fell to 26 Nm, and after Wallis stabilization, it was restored to 46 Nm. The stiffness of the intact overlying adjacent segment (L3–L4) was significantly higher above the Wallis than above the pedicle screw-augmented PLIF (46 Nm vs. 35 Nm; p < 0.05), confirming their own experimental cadaver studies [73, 74]. Other cadaver studies of the second-generation Wallis have shown a 14 % reduction in flexion-extension ROM of the stabilized L3–L4 segment, with small increases in ROM at the uninstrumented L2–L3 and L4–L5 segments of 7 % and 3.5 %, respectively, and little influence on lateral bending and axial rotation [75]. Similar biomechanical results in cadavers have been found with another device that has a spacer and a tension band, which device also was shown to reduce pressures in the posterior annulus and central nucleus [22]. Comparing Wallis to a pedicle screw-based dynamic stabilization system, Schulte et al. reported that the Wallis implant reduced extension by 69 % and flexion by 62 %, with almost no action on lateral bending or axial rotation [76].

A study of distance between spinous processes showed that the variations in interspinous process distance (ISPD) are greater in patients with degenerative disk disease than in healthy subjects, demonstrating the risk of implant dislocation for interspinous devices that do not limit ISPD during flexion [77]. Indeed, among dynamic stabilization devices that do not strongly limit flexion, dislocation has been reported for Coflex in several studies [12, 78–80]. In this respect, the DIAM device appears to be an exception. Even though cadaver studies show that it fails to limit flexion [66, 67], the tethering laces probably explain the lack of reported postoperative dislocations [81]. Next to the Wallis system, other interspinous stabilization devices with strong flexion-limiting bands have been developed to restore segmental stiffness more consistently, including the Ligament Vertebral de Renfort [16], Dallos [18], InSWing [20], and Locker [21]. In a porcine study, the supraspinous ligament (SSL) and the laces of DIAM were both observed to have a mechanical role, leading the authors to recommend preservation of the SSL and use of the laces, which also were thought to prevent postoperative dislocation [82]. The LimiFlex device has flexion-limiting bands, but not an interspinous spacer [3].