The rationale behind arthrodesis of a selected lumbar spinal segment is to reduce the nociceptive load by wide removal of the disrupted disc material and the stabilization of the affected motion segment(s) that are thought to be causing pain due to continued motion of the involved sensitized levels. Broadly, as a technique, it has long been considered the "gold standard" for LDDD regardless of the approach employed due to its relatively reliable results in properly selected patients. There are three main fusion techniques currently employed, namely posterolateral fusion, interbody fusion and combined interbody fusion with posterolateral fusion, also known as a 360 degree or circumferential fusion.

Posterolateral fusion (PLF) targets only the posterior elements. Through a midline skin incision with either a posterior approach or bilateral muscle splitting fascial incisions, facets joints and transverse processes bilaterally are decorticated. Local or autologous bone graft from the iliac crest can be packed into the posterolateral gutters to promote fusion. Historically, posterolateral fusion was performed without instrumentation, but due to the relatively high non-fusion rate [4849], pedicle screws instrumentation has become standard. Fusion rates have in fact been reported as ranging from 46 to 100% and the pseudoarthrosis rate have been reported as high as 32% for non-instrumented fusion [26]. Current evidence supports the use of instrumentation to ultimately increase fusion rates [4849]. However, the evidence of an improved functional outcome for an instrumented fusion instead of a non-instrumented one remains very limited [4849]. A few small clinical trials have shown improved clinical functional outcomes and better fusion rate amongst instrumented patients compared to those with non-instrumented fusion [4753]. However, other studies have suggested no clinical benefit [4854]. It should be noted that not all pseudoarthrosis results may be painful or unstable [49] and therefore, they may not necessarily warrant revision surgery.

In search of better outcomes, spinal surgeons have switched their attention back to the disc, postulating it as the main pain generator. Weatherley et al. [55] investigated a subgroup of patients who had solid posterior fusion with persistent LBP reproduced by provocative discography. These patients experienced relief of the symptoms following subtotal disc clearance and anterior interbody fusion demonstrating the importance of disc clearance and interbody fusion.

Interbody fusion can be performed with either a posterior or anterior approach. Interbody fusion techniques offer several biologic and biomechanical advantages over posterolateral fusion: (1) They address the disc as the pain generator in patients with discogenic pain. (2) The anterior column of the spine supports 80% of the body load and consequently interbody devices are subjected to compressive load. Together with a large surface area, this biomechanical advantage facilitates fusion. (3) Insertion of interbody devices allows restoration of intervertebral height, correction of lumbar lordosis and restoration of sagittal balance depending on design and final positioning on the endplates, with potential for indirect decompression of the neural elements.

Various approaches have been described to achieve this purpose: (1) Posterior lumbar interbody fusion (PLIF) accesses both columns (anterior and posterior) through a single posterior approach and involves a wide laminectomy, with partial or complete facetectomy, subtotal discectomy, endplate preparation and insertion of interbody device with or without supplemental pedicle screw instrumentation. This approach allows excellent spinal canal decompression, which can address concurrent central stenosis. The wider the resection, the lesser the neural retraction to access the diseased intervertebral disc, although this approach may destabilize the segment and it is recommended that in order to achieve stability of the construct, pedicle screws be employed especially above L5–S1 and when multiple levels are involved [56]. The disc is approached through the axilla formed by the dural sac and exiting upper nerve root. The interbody device consists of either a cage (typically manufactured from Peek and carbon fibre) or a bone graft spacer, which is then inserted after end plate preparation with curettes and rongeurs. PLIF can also be supplemented with posterolateral fusion to increase fusion rates. (2) Transforaminal lumbar interbody fusion (TLIF) is a modification of the PLIF technique where the intervertebral disc is exposed unilaterally through a transforaminal approach with subtotal facetectomy in conjunction with pedicle screw instrumentation (Fig. 4). Nerve root retraction is required to access the disc, exposing the patient to a risk of nerve injury and radiculitis. Bilateral pedicle screw instrumentation is recommended to restore stability. Similar to PLIF, contralateral posterolateral fusion can be added to increase fusion rates. While some advocates of this procedure maintain that they can treat a wide spectrum of pathologies through this approach, including degenerative spondylolisthesis, a potential disadvantage is that it is a more demanding technique for implantation of the interbody device, which can affect the ability to adequately achieve sufficient lordosis and subtotal disc clearance. (3) The anterior approach for lumbar interbody fusion (ALIF) is performed through either a transperitoneal or retroperitoneal approach. This approach allows access to the lumbar discs from L2 to the sacrum. Renal vessels and the diaphragm limit the more proximal extension of the exposure. The pubic symphysis occasionally also precludes appropriate visualisation and instrumentation of the L5–S1 disc space in patients with a deep-seated L5–S1 level relative to the pelvis. Lateral radiographs including the pubis are obtained during the preoperative planning stages to assess suitability of this approach at that level. Preparation of end plates and insertion of graft or implant are similar to PLIF and TLIF but the direct front-on approach allows better visualization and larger size of implants with greater capacity to increase disc height and lordosis.

The minimally invasive surgery (MIS) fusion techniques have been developed that allow access while limiting soft tissue dissection and muscle damage. These are often modifications of pre-existing open techniques. This is in response to higher patient demand and expectation for faster recovery times in line with minimally invasive technique advances in other areas of surgery in the last three decades. Proponents of MIS interbody fusion techniques have also highlighted reduced intraoperative blood loss and muscle damage [5758] as well as faster return to work and decreased opioid dependence for the patient [59].

MIS adaptation of TLIF utilises the Wiltse paraspinal approach [60] through multifidus and longissimus intermuscular plane for unilateral facetectomy and pedicle screw instrumentation. This technique reduces muscle stripping, preserves blood supply and muscular attachment and thereby reduces erector muscle dysfunction and fibrosis [61].

More recently lateral lumbar interbody fusion and anterior to psoas approaches have been developed, which are MIS or mini-open in nature. The former technique gains access to the lateral lumbar spine through a transpsoas approach while the latter occurs anterior to psoas [62]. The patient is placed in a lateral decubitus and a small lateral incision is made to gain access through the retroperitoneal space. Neural monitoring is then placed in the posts to identify the lumbosacral plexus. Safe access to the lumbar spine is usually obtained within the anterior one third to one half of the psoas; however, neurologic adverse events overall are reported for 18%–40% of patients [636465].

Over the past few decades, there has been an explosion of different instrumentation types, fusion methods and bone graft sources developed to improve fusion rates and improve functional outcomes. Many studies have since been published to establish their utility, but results are often contradictory, with inclusion of a wide range of clinical diagnoses such as isthmic and degenerative spondylolisthesis and differing methods of patient selection that are often uncontrolled for comorbidities such as smoking, diabetes mellitus and steroid use. This makes comparisons difficult when attempting to advocate for one technique over another.

With regards to ALIF, the advantages of include preservation of posterior tension-band functioning musculature, improved access to the intervertebral space for disc removal and precise placement of interbody graft or implant for ease of height and lordosis restoration. The approach has several other advantages such as obviating the need for prone positioning, reduced blood loss (assuming no great vessel injury occurring intraoperatively), reduced neurological injury and adjacent segment disease [6667]. ALIF also allows insertion of a larger structural graft without endangering neural structures. However, it can be technically demanding, particularly in the case of an obese patient and those with widespread atherosclerosis, and requires confidence with dealing with potential vascular complications due to retraction of vessels with a potential for tears to the common iliac and iliolumbar vessels [68]. Furthermore, injury to the parasympathethic plexus may result in retrograde ejaculation and autonomic sensory disturbances in the lower limbs. These disadvantages have led to a movement of surgeons employing the expertise of "access" vascular surgeons and intraoperative heparin administration to facilitate access to the intervertebral disc in the hope of reducing complications for the patient [686970]. Bateman et al. [71] undertook a systematic review of the complication of anterior approach surgery finding intraoperative and postoperative complication rates of 9.1% and 5.2%, respectively. The most common complications reported were venous injury (3.2%), retrograde ejaculation (2.7%), neurologic injury (2%), prosthesis related (2%), postoperative ileus (1.4%), superficial infection (1%), and others (1.3%). In addition, should an ALIF technique be considered for a candidate with poorer bone stock or previous wide laminectomies, this would be a relative indication for supplemental posterior instrumentation or posterolateral fusion compared to a standalone procedure.

Early experience with ALIF as a standalone procedure was associated with high rates of non-union [72], but reports on newer techniques and devices designed for ALIF have shown good clinical outcomes and comparable fusion rates with other techniques [73] as well as comparable biomechanical properties to the addition of posterior instrumentation (Fig. 5) [74]. ALIF combined with posterolateral-instrumented fusion or also known as circumferential fusion provides the best fusion rates; however, this involves intraoperative repositioning of the patient and the added morbidity of increased operative time.

The main theoretical benefit of PLIF and TLIF over ALIF is the significant lower risk of injury to anterior vascular elements. PLIF also allows for simultaneous decompression of central canal and bilateral foraminae. While TLIF allows for only unilateral decompression of foramina compared to PLIF, it may lead to a lower incidence of neural complications (dural tears and epidural fibrosis because of the minimal and unilateral exposure of the cauda equina). PLIF is not recommended for cases involving the upper lumbar level (L1–2) due to the need to manipulate the conus medullaris. While TLIF is considered by some authors as safer to PLIF at this level due to lesser neural retraction, great caution is still required. In a meta-analysis comparing PLIF to TLIF, PLIF was associated with a significantly higher complication rates [75], in particular dural tears and root injury as well as graft malposition. However, there were no differences with regards to clinical satisfaction, blood loss and radiographic evidence of fusion between the two groups, even if PLIF required longer operative times.

Unilateral instrumentation has also been advocated as an alternative to bilateral instrumentation for spinal fusion with interbody device. The theoretical advantage is shorter operative times and lower operative costs for implants [76]. In a systematic review of ten randomized trials, Molinari et al. [76] reported no difference in clinical outcomes, union rates nor complications when single-level unilateral instrumentation was used for degenerative conditions in the lumbar spine compared to bilateral instrumented fusion with PLIF and TLIF.

When comparing the three main types of fusion techniques—posterolateral, posterolateral with instrumentation, and circumferential—SLSSG found that circumferential fusion had the highest fusion rate, followed by instrumented posterolateral fusion, for all groups having significantly reduced pain and disability and no overall significant differences in clinical outcomes between the groups. The circumferential group, however, had significantly more complications, prolonged operating times and prolonged hospital stays.

The issue of fusion rate needs careful interpretation. Radiological determination of lumbar fusion is known to have poor intra- and inter-observer consistency and depends on the method utilised. Sensitivity and specificity of plain radiographs have been reported to be as low as 43% and 50%, respectively [48]. Additionally reported fusion rates in the literature can be flawed with bias, as in many cases the designation of fusion status is assigned by the operating surgeon. In addition, fusion rates are not the only measure of success after lumbar spine fusion and indeed some pseudoarthroses (such as the "locked" pseudoarthrosis [77]) are not incompatible with good outcomes.

A randomized controlled study by Videbaek et al. [78] was in contrast to the findings of the Swedish Lumbar Spine Study Group, and reported improved mid-term outcomes with circumferential fusion when compared to instrumented posterolateral fusion alone. In this trial, there was a statistically significant improvement in pain score and functional score with a 5 to 9 year follow up [78]. The study population was significantly younger when compared to the Swedish group with a majority of patients being at work. A further study by Soegaard et al. [79] supported these results, reporting superior cost saving in the circumferential group compared with posterolateral fusion group in the long term.

In line with current evidence, one may suggest choosing interbody fusion techniques in the younger, more active and working individuals and in patients with an increased risk of non-union i.e. smokers. Also, the debate between conventional or minimally invasive techniques will continue until long-term comparisons of these techniques in well-designed studies are available. In order to displace conventional techniques as the standard of care, MIS techniques would need to demonstrate at least equivalent outcomes without additional costs and increased complications, in addition to the currently reported beneficial short term outcomes of reduced hospital time and quicker return to employment [80].

Previously, tricortical blocks of autologous iliac crest bone graft were used in surgery. This presented substantial technical challenges, including the sizing of the grafts, its stability once impacted into the intervertebral disc space and the issue of subsequent collapse and subsidence. The development of interbody devices with instruments designed for ease of introduction has increased the technical ease of interbody fusion. There is currently a wide range of materials used as interbody devices with which various graft options can be used to enhance fusion. These interbody devices also have design characteristics that may help in restoring lumbar lordosis, sagittal balance, indirect neuroforaminal decompression and physiological load through the anterior spinal column. Titanium cages have over the years been popular but can be associated with subsidence. Interbody cages made of carbon fibre and the non-absorbable, biocompatible material polyetheretherketone (PEEK) are radiolucent and are thought to have a modulus of elasticity closer to bone. These cages may have lower subsidence rates and, in addition, their usage allows easier assessment of fusion rates on radiographs. However, some debate exists on the enhancement by titanium implants for bone incorporation into the implant as opposed to PEEK, which is thought to stimulate fibrous, non-mineralised tissue.

Iliac crest graft has been long been considered the gold standard for spinal fusion, but it has several disadvantages compared to currently available alternatives. These include associated donor site morbidity, increased blood loss, prolonged hospital stay and often limited graft volume in older patients. Bone graft substitutes can be broadly classified into two categories: (1) There are osteoinductive agents that induce formation of bone, and these include agents such as recombinant human bone morphogenetic protein, bone marrow aspirate etc. (2) There are also osteoconductive agents, including demineralized bone matrix, tricalcium phosphate etc., that provide a scaffold for new bone growth.

Allograft yields comparable fusion rates to autograft without donor site morbidity [81], but with an increased cost and a potential for infectious disease transmission. Demineralised bone matrix (DBM), while still on-going investigation regarding its efficacy, has had early positive reports with high fusion rates [82]. Currently DBM is used as mass extender, combined with local autologous bone graft, rather than used in isolation. Bone morphogenetic proteins (BMPs) are reported to be at least as effective as autologous bone graft substitutes or extenders [83]; however, there are concerns regarding their safety. With BMPs, extensive osteolysis resulting in graft subsidence and cage migration has been reported [84]. Higher rates of retrograde ejaculation have also been reported when used in ALIF, thought to be due to the inflammatory response associated with release of cytokines due to BMP in closed vicinity of the parasympathetic plexus [85]. However, Scott-Young [86] reported the incidence of the above to be only 0.7% in a large cohort of males who underwent ALIF and similarly another study did not find the high rates of complication previously reported [87]. Although not proven, there are theoretical concerns for a cancer when applying exogenous transforming growth factors such as BMPs [88]. A multitude of other graft substitutes and biological mediators are being introduced into clinical practice, but are beyond the scope of this review.

Lumbar disc arthroplasty (LDA) has been increasingly adopted into routine spinal practice since its introduction in 1966 [89]. After a lag period, the Charité Hospital in Berlin, East Germany led the trend for widespread implantation beginning in the 1980s. LDA typically is performed through an anterior approach at lower lumbar levels (L4/5 and L5/S1) in younger and active individuals with limited ASD, adequate bone stock, absence of spondylolisthesis and normal facet joints (Fig. 6). Non-inferiority studies involving both cohorts of patients implanted with the Charité prosthesis compared to ALIF [90], the ProDisc prosthesis compared against circumferential fusion [91] and the Maverick compared to ALIF [92], respectively, showed favourable results for LDA but each had methodological flaws. There is no current conclusive evidence of LDA superiority in the long-term in level I studies and the surgery is technically more challenging. However, mid-term studies of LDA have reported satisfactory clinical results and implant survival and comparable complication profiles to fusion [9394] with proponents of LDA supporting its use for several theoretical and clinical advantages.

The major driver for development of this technique has been the motion preserving philosophy as an alternative to fusion with theoretically lower risk for ASD. European studies have cited rates of ASD at 10-year follow-up of LDA to be 2.0%–2.8% and up to 4.5% in a large North American study [95], which compares favourably to historical rates in lumbar fusion of between 14 and 29%. However, it may be unfair to compare ASD to fusion procedures indiscriminately as the anterior approach itself has been reported as being associated with lower rates of ASD [9697].

Other advantages cited for LDA include having no need either for grafting or screw based fixation [98] and having a lower reoperation rate with at least equivalent rates of return to work and clinical outcomes to fusion at mid-term [99]. In one study, David [100] concluded that the "rate of reoperation secondary to adjacent segment disease is ten times lower than the rates reported in the literature for fusion." Furthermore, the rates of secondary sacroiliac joint dysfunction following LDA were low in the mid-term [101]. Studies of modern LDAs have also reported faster patient recovery, higher satisfaction in the short-term compared to the fusion techniques [95]. Patient satisfaction rates have been in the order of 88% or more [90102103], at least equivalent to fusion satisfaction rates quoted as 85%–95% [104105], although some series have reported a mild but statistically significant deterioration in outcomes of patients transitioning from early to mid-term follow-up [94]. High rates of return to work and lower long-term unemployment have also been reported as other benefits for the LDAs [90]. These results may also be further improved with focused physiotherapy [106].

Disadvantages cited in the literature are the disappointing outcomes from some of the earlier designs (perhaps confounded also by patient selection), the lack of long-term outcomes of modern designs, the lower versatility in indications compared to fusion and the difficulty in acquiring adequate surgical expertise in performing this procedure. Many of the cited complications such as implant extrusion and vascular injury can be linked to inadequate training, improper sizing and lack of confirmation of satisfactory placement on imaging [107] with implant impingement linked to the aforementioned factors [108]. Design variables relating to constraint and core mobility have also been shown to alter wear and force transmission in in vitro studies [109,110].

Earlier, some of the initial LDA designs led to inconsistent outcomes [111]. For example, the cumulative survival for the Acroflex LDA was only 60.7% at 10 years when first revision surgery was used as the end point. Etiology of the implant failure reported by authors [112] included failure of osteointergration, midsubstance elastomeric tears and osteolysis. Complication rates, while reported to be lower than fusion in the Charité trials (29.1% for arthroplasty and 50.2% for fusion at 2-year follow-up), are still significant [113]. These can be broadly separated into those related to the anterior surgical approach (e.g., vascular injury, nerve root injury, retrograde ejaculation), prosthesis/fusion failure (e.g., subsidence, osteolysis, migration, implant fracture, endplate fracture, pseudoarthrosis), heterotopic ossification (up to 76% at 3 years) and subsequent hypomobility of the implant and donorsite complications [93]. Device failures necessitating repeat operations have been reported at 5.4% to 6.3%. Salvage procedures involving conversion to spinal fusion are technically demanding; however, they appear to improve outcomes modestly in failed LDA [112]. Earlier designs were also associated with a high rate, in the order of 60%, of spontaneous fusion in the long term [111]. Metal ion and polyethylene wear particle release from the articulation interface also add to the disadvantages [114]. Limited versatility exists compared to fusion techniques as there are relative contraindications for three or more lumbar levels, high lumbar disease (above L3), spondylolisthesis, severe ligamentous instability, facet degeneration, adjacent prior fusion and conditions affecting bony healing (e.g. rheumatoid arthritis, diabetes mellitus and steroid use) [115116].

Stem cell therapy combined with tissue engineering approaches for effective delivery into the degenerative disc has been proposed as a minimally invasive alternative to manage LBP. This therapy focuses on the restoration of extracellular matrix of the "damaged" disc where autologous nucleus pulposus cells or bone marrow derived mesenchymal stem cells are harvested, expanded in vitro and delivered into the nucleus pulposus, with the hope of stabilizing or reversing LDDD pathology. In a recent systematic review evaluating the effectiveness of a tissue engineered approach in the treatment of DDD in controlled animal studies, the technique has shown promising results in improvement in disc height and MRI signal intensity but none of the studies showed restoration to the properties of a healthy disc [127]. Human trials on the use of mesenchymal stem cells from autologous bone marrow in patients with LDDD are currently on-going in multiple centres. Finally, short term results of intradiscal pulsed radiofrequency have been promising and have shown not to be significantly different from intradiscal electrothermal therapy at 6 months [128]. Results with appropriate follow-up are not yet available to indicate if such techniques are safe, cost-effective and reliable in the long-term.