3.4.3.3.1. Neurological Function

There was inadequate evidence from one small RCT on the comparative effectiveness of ACDF, physiotherapy and treatment with a cervical collar on neurologic function in patients with cervico-brachila pain without spinal cord compression (SOE: Insufficient).

Specific muscle strength before and after treatment was also assessed.²⁶ Patients in the surgery group experienced greater improvements in muscle strength (strength expressed as the ratio of the affected to the unaffected side) at 14 to 16 weeks in pinch grip, elbow extension and shoulder internal rotation compared with patients treated with physiotherapy and greater improvements in wrist flexion and elbow flexion compared to those treated with a cervical collar (data not provided). At 16 months, patients treated with surgery experienced greater improvements in wrist extension, elbow extension, shoulder abduction, and shoulder internal rotation compared with patients treated with physiotherapy. There were no differences in strength improvement between surgery and collar treatment or between physiotherapy and collar treatment at 16 months (data not provided).

At 14 to 16 weeks posttreatment, there was no difference in the likelihood of improvement in paresthesias with surgery compared with physiotherapy or collar treatment (N=81, 52% vs. 45% vs. 37%, p>0.05) but a large increase in the likelihood of improvement in sensory loss with surgery compared with either treatment (41% vs. 15%, RR 2.75, 95% CI 1.0 to 7.5, both comparisons with surgery).²⁶ At 16 months, there remained no difference between treatment in the likelihood of improvement in paresthesias between surgery, physiotherapy, and treatment with a collar (N=81, 58% vs. 67% vs. 66%, p>0.05). There was also no difference between treatments in the likelihood of improvement in sensory loss at 16 months (N=81, 27% vs. 14% vs. 15%, p>0.05).

3.4.3.3.2. General Function

There was inadequate evidence from one small RCT on the comparative effectiveness of ACDF, physiotherapy and treatment with a cervical collar on general function in patients with cervico-brachila pain without spinal cord compression (SOE: Insufficient).

The ability to complete basic activities of daily life (e.g., dressing, prolonged sitting) to more rigorous physical activity (e.g., running, heavy work) was assessed using the disability rating index (DRI).²⁵ Overall mean score on the DRI ranges from 0 to 100, with ability on each of 12 activities rated using a 0-100 VAS scale indicating “without difficulty” to “not at all.” There was no difference between treatment with surgery versus physiotherapy at 14-16 weeks on improvement in disability, however treatment with surgery resulted in improved dressing and heavy work compared with treatment with a collar, while treatment with physiotherapy was associated with greater ability to walk, sit for a long time, and complete heavy work compared with collar treatment (p<0.05, data not provided). At 16 months the ability to do heavy work was greater with surgery compared to the other treatments (p<0.05, data not provided). No other differences on the DRI were noted.

Although findings from this small study tended to favor surgery, especially in the short term, these findings should be interpreted with caution due to patients receiving additional treatments beyond the randomized treatment and the heterogeneity of treatment (especially physiotherapy). After 16 weeks, 8/27 surgery patients (30%) underwent a second surgery. Additionally, one patient treated with physiotherapy (4%) and five treated with collar (19%) underwent surgery. Forty-one percent of surgery patients (11/27) received physiotherapy as did 44% (12/27) of patients treated with a collar. Additionally, the use of specific physiotherapy modalities (e.g., traction, exercises, cryotherapy) varied and was at the discretion of the local physiotherapist.

3.5.3.1.1. Fusion

No study reported fusion outcomes.

3.5.3.1.2. Pain

There was no difference in pain between the use of a post-operative collar plus laminoplasty versus laminoplasty alone (SOE: Low). There was inadequate evidence to determine the effects on pain with laminoplasty plus exercise versus laminoplasty alone (SOE: Insufficient).

Single-door laminoplasty plus rigid Philadelphia collar worn for 3 weeks post-operatively was associated with less improvement in mean VAS scores (0-10 scale) than laminoplasty alone at weeks 1 (0.8 vs. 3.8, p=0.023) and 2 (−0.9 vs. 1.8, p=0.046) in one trial rated low risk of bias (N=35), with no difference at other timepoints (3 weeks: −1.2 vs. 1.1, p=0.148) or at other followup times (6 weeks and 3, 6, and 12 months).²⁸ One trial (N=90) compared modified double-door laminoplasty plus Philadelphia collar worn for 2 weeks post-operatively and found no differences in change in VAS (0-10 scale) at 12 months (−0.19 vs. −0.04, p>0.05) or throughout the study period (p=0.487).³⁰

One RCT (N=65) found no difference in mean VAS scores (0-100 scale) for neck pain and stiffness at 2 weeks and 3 months postoperative between muscle-preserving laminoplasty with exercises versus laminoplasty alone (3 months: −1.8 vs. −2.5, p=0.623).³¹

3.5.3.1.3. Function

3.5.3.1.4. Quality of Life

No study reported quality of life outcomes.

3.5.3.2.1. Fusion

There was inadequate evidence to determine the effects on fusion between ACDF with or without collar use (SOE: Insufficient). Use of post-operative PEMF stimulation after ACDF was associated with increased fusion versus treatment with ACDF alone (SOE: Low).

All ACDF patients in one 24-month trial (N=33) achieved radiographic fusion regardless of collar use (100% vs. 100%).²⁷ Surgical details were not provided.

PEMF was associated with small increase in fusion rates at 6 months in one trial (N=323) based on a per protocol analysis versus ACDF with no PEMF (N=240; 83.6% vs. 68.6%, p=0.0065); fusion rates were also improved in intent-to-treat analyses assuming missing patients fused (N=323; 85.9% vs. 76.3%, p=0.0269) or imputing patient status at last visit (N=281; 78.2% vs. 64.8%, p=0.0127), but not when assuming missing patients did not fuse (65.6% vs. 56.3%, p=0.0835).²⁹ However, there was no difference in fusion rates in the per protocol analysis at 12 months.²⁹ This study used a Smith-Robinson technique with allograft and cervical plate system.

3.5.3.2.2. Pain

The ACDF trial of PEMF versus no PEMF found similar VAS scores for shoulder/arm pain at rest or with activity at 6 and 12 months postoperative (date provided in graph form)²⁹ (SOE: Low).

3.5.3.2.3. Function

3.5.3.2.4. Quality of Life

No study reported quality of life outcomes.

3.6.3.1.1. Fusion

There was inadequate evidence to determine benefits and harms of anterior versus posterior surgical approaches on cervical fusion (SOE: Insufficient).

One RCT (N= 30) rated high risk of bias reported that no participants in either the ACF group or the posterior cervical foraminotomy group had radiologic evidence of instability on cervical x-rays at time of discharge or at a mean of 14 months.³² Authors did not define stability or criteria for determining fusion.

3.6.3.1.2. Pain

There were no differences in neck and arm pain between anterior versus posterior approaches in the short (3, 6 months) and intermediate term (12, 24 months) (SOE: Low); there was inadequate evidence to determine the benefits and harms of anterior versus posterior approaches on neck pain immediately post-operative (SOE: Insufficient).

At 12 months the proportion of ACDF vs. PCF patients experiencing a 26-point improvement (0-100 scale) in VAS neck pain (62% vs. 52%) or 41-point improvement in VAS neck pain (60% vs. 54%) was reported as comparable in one RCT (N=243);³⁵ an RR could not be calculated.

One small trial (N=30) rated high risk of bias reported that ACF was associated with lower neck pain VAS scores (0-10 scale) within a week of discharge (p<0.001), however the reported confidence interval for the difference between groups suggested no difference (MD −3.13, 95% CI −4.52 to 1.75) and is likely a typographical error and should be (MD −3.13, 95% CI −4.52 to −1.75).³² One RCT (N=175) also rated high risk of bias, compared ACDF versus PCF at 3, 6, 12, and 24 months for arm pain VAS (0-100 scale), neck pain VAS (0-100 scale) and North American Spine Society (NASS) pain (0-6 scale).³³ The mean differences across measures did not change with time and there were no differences between ACDF and PCF in arm pain VAS (range from −1 to 1), neck pain VAS scores (range from 1 to 4) or NASS pain scores (range from −0.1 to 0.1) at any timepoint. Statistical tests were not reported and reported data were inadequate to calculate confidence intervals for effect sizes, but the authors noted that the clinical results were the same in both groups. The largest RCT (n=243, moderate quality) found no difference in VAS neck pain scores at 12 months (MD −2.70, 95% CI −9.67 to 4.27) or VAS arm pain (MD – 2.80, 95% CI, −8.85 to 3.25).³⁵ Pooled estimates across the RCTs also reveal no difference in VAS arm pain at 12 months between ACDF and PCF (2 RCTs, N=403, MD −1.36, 95% CI −5.23 to 1.86, I²= 0%; Appendix F Figure F-1).^33,35 Across the same two RCTs, there was again no difference between ACDF and PCF in VAS neck pain (MD 0.31, 95% CI −6.20 to 5.81, I²=10.6%; Appendix F Figure F-2).^33,35

The fourth RCT (N=72) rated moderate risk of bias, reported similar rates of patient-reported complete or partial pain improvement (unvalidated measure) for anterior approaches (ACD and ACDF) versus PCF at day 1 postoperatively (100% vs. 100%, RR 1.00), at 2 months (98% vs. 100%, RR 0.98, 95% CI 0.94 to 1.02, p=0.32), and at approximately 60 months postoperatively (96.5% vs. 100%, RR 0.96, 95% CI 0.90 to 1.03, p=0.32).³⁴

Findings for pain from two NRSIs were consistent with those of the RCTs. The larger study (N=688) found no difference in mean scores for VAS arm pain (0-10 scale) at 3 months (4.20 vs. 3.82, MD 0.38, p>0.05), 12 months (4.06 vs. 4.07, MD 0.01, p>0.05) or 24 months (3.85 vs. 4.48, MD −0.63, p>0.05).³⁸ In the smaller NRSI (N=70) rated high risk of bias, there were no differences between ACDF versus PCF in VAS score (0-10 scale, not specified for arm or neck pain, 2.6 vs. 3.0, MD −0.4, p =0.04) at 12 months.³⁶ Reported estimates appear to be unadjusted.

3.6.3.1.3. Function

3.6.3.1.4. Quality of Life

There was no difference in EuroQOL-5 Dimensions (EQ-5D, scale 0-1) at 12 months between ACDF and PCF in one RCT (SOE: Low).³⁵

The RCT (N=243) found no difference on EQ-5D between ACDF and PCF in either the proportion of patient meeting a clinically important difference of 0.24 improvement (38% vs. 38%) or in change scores at 12 months (MD −0.01, 95% CI −0.06 to 0.10).³⁵ Similarly, one NRSI (N=70) rated high risk of bias found no difference in EuroQOL-5 Dimensions (EQ-5D, scale 0-1) at 12 months for ACDF (MD 0.69, 95% CI 0.61 to 0.77) versus PCF (MD 0.72, 95% CI 0.64 to 0.80, p=0.60).³⁶

3.6.3.1.5. Reoperation

There was no difference in the likelihood of reoperation between anterior and posterior procedures across four RCTs^32–35 (2 of which were rated high risk of bias) or in one retrospective NRSI (N=328)³⁷ (Figure 3) (SOE: Low). Exclusion of the high risk of bias RCTs did not substantially change the estimate (2 RCTs, RR 0.70, 95% CI 0.30 to 1.61, I²= 0%).^34,35

Figure 3

Reoperation: anterior versus posterior cervical foraminotomy. ACF = anterior cervical foraminotomy, ACD = anterior cervical decompression without fusion; ACDF = anterior cervical discectomy and fusion; CI = confidence interval; PL = profile likelihood. (more...)

3.6.3.1.6. Harms

There were no differences in neurologic deficits between anterior and posterior approaches, although results were reported inconsistently (SOE: Low); reporting of other adverse events was limited (SOE: Insufficient).

Description and reporting of serious adverse events was limited. One RCT (N=243) reported similar rates of surgery-related adverse events for ACDF and PCF 6% in both groups).³⁵ Serious (not specified as surgery related) included: post-operative events (anaphylactic reaction to antibiotics, n=1, wound hematoma not requiring surgery, n=1, pulmonary embolism, n=1) and events requiring hospitalization (wound problems 0.8% vs. 1.7%, cardio-thoracic problems 0.08% vs. 2.5%). Slight cage subsidence was reported in one ACDF patient but there were no complaints and no reoperation was required.

Three RCTS (2 of which were rated high risk of bias) reported that no serious adverse events occurred for any patients.^32–34 One RCT (N=72) that compared ACD and ACDF to PCF reported zero deaths.³⁴ One propensity score matched NRSI (N=46,598) reported higher 30-day mortality with ACDF versus PCF (MD 1 event per 10,000 cases, 95% CI 0.0 to 2.0 per 10,000 cases, p=0.012).³⁹ Although the MD is significant, it is small, suggesting the possibility of 0 to 2 deaths with PCF. Given that administrative data are subject to misclassification and potential for inadequate adjustment for confounders, this finding should be interpreted cautiously.

Neurologic deficits were reported inconsistently across studies. In one RCT (N=243) there was no difference in new radicular symptoms between ACDF and PCF recipients (3.2% vs. 0.8%, RR 3.84, 95% CI 0.43 to 33.85) as were persistent radicular symptoms (1.6% vs. 6.7%, RR 0.24, 95% CI 0.05 to 1.11); estimates are imprecise.³⁵ One RCT (N=72) found no difference in anterior versus posterior approaches for new weakness (8% vs. 14%, RR 0.59, 95% CI 0.14 to 2.40, p=0.46) or new numbness (6% vs. 9%, RR 0.66, 95% CI 0.12 to 3.68, p=0.63).³⁴ The other two RCTs reported specific neurologic deficits: in one small trial (N=30) no patients in either group developed Horner’s syndrome;³² the other trial (N=175) reported that no patients experienced damage to myelin resulting in paralysis of any degree.³³ One NRSI (N=70) reported that one patient who underwent PCF experienced C6 nerve injury, but did not provide data for patients who underwent ACDF.³⁶ Central nervous system complications at 30 days postoperatively was similar between anterior and posterior procedures in a large NRSI (N=46,598, MD 4 per 10,000, 95% CI −14 to 22 per 10,000).³⁹

Dysphagia was reported inconsistently across studies. One RCT (N=243), reported one case of unresolved dysphagia at 12 months in the ACDF group.³⁵ One RCT (N=175) reported transient difficulty swallowing for three patients who underwent ACDF and no patients who underwent PCF.³³ In a propensity score matched NRSI (N=46,598), ACDF was associated with higher rates of dysphagia/dysphonia at 30 days versus PCF (MD 14.5 per 1,000 cases, 95% CI 12.6 to 16.4 per 1000, p<0.001).³⁹ Neither study provided information on severity of dysphagia or need for intervention.

One large NRSI (N=46,598) reported that the following were rare but more common with ACDF versus PCF within 30 days after surgery: vascular injury (MD 2 per 10,000 cases, 95% CI 1 to 3 per 10,000 cases, p=0.001), cerebrospinal fluid leak (MD 2 per 10,000 cases, 95% CI 1 to 3 per 10,000 patients, p=0.002), and deep venous thrombus (9 per 10,000 cases, 95% CI 2 to 16 per 10,000 patients, p=0.01). There were no differences between anterior and posterior approaches for pulmonary embolism (MD 2 per 10,000, 95% CI −9 to 12 per 10,000 cases).³⁹

3.7.3.3.1. Neurologic Function

There was low-strength evidence of no difference in neurologic function between anterior and posterior approaches in participants with three or more level disease in the intermediate term (SOE: Low); there was inadequate evidence for determining the benefits and harms on neurologic function in the short term (SOE: Insufficient).

There was no difference in neurologic function at intermediate term (12 months) in one small RCT rated high risk of bias (N=32, MD 0.28, 95% CI −0.41 to 0.98, Japanese Orthopaedic Association Scale [JOA] scores, 0-18 scale)⁴⁰ and two NRSIs rated moderate risk of bias (N=506, MD 0.15, 95% CI −0.29 to 0.60, I²=74.0%, mJOA scores, 0-18 scale)^40,41,44 that compared ACDF with posterior laminoplasty (RCT) or PCDF (NRSIs) for 3- to 5-level disease (Figure 4) (SOE: Low). There was also no difference between groups in JOA scores short term in the RCT (N=32): 3 months (MD −0.40, 95% CI −1.76 to 0.96) and 6 months (MD 0.20, 95% CI −1.14 to 1.54).⁴⁰ The pooled estimate across the two NRSIs had substantial heterogeneity (Figure 4), which may be due in part to different study designs, variables controlled for in multivariate analyses, and types of posterior procedures used. The prospective NRSI⁴⁴ showed no difference between groups and included patients who underwent laminoplasty (14%) (all others had PCDF); it was unclear which baseline confounders were controlled for in this study. The retrospective NRSI⁴¹ showed a large improvement with ACDF versus PCDF approaches; multivariate logistic regression models controlled for 19 different baseline variables.

Figure 4

Neurologic function (JOA or mJOA scores): anterior versus posterior approaches for ≥3 levels. ACDF = anterior cervical discectomy and fusion; CI = confidence interval; CSM = cervical spondylotic myelopathy; JOA = Japanese Orthopaedic Association; (more...)

One prospective NRSI (N=264) assessed neurologic function with the Nurick score (0-5 scale) and found no difference between 3- to 5-level ACDF and posterior approaches (laminectomy and fusion [86%] or laminoplasty [14%]) in mean change from baseline to 12 months after adjusting for baseline characteristics (MD in change scores 0.19, 95% CI −0.20 to 0.58⁴⁴).

3.7.3.3.2. General Function

There were no differences between anterior and posterior surgery for 3- to 5-level disease at intermediate term (12 months) for any function measure reported across two NRSIs (N=509)^41,44 (SOE: Low). One prospective NRSI (N=264) compared ACDF with laminectomy and fusion (86%) or laminoplasty (14%) and reported the change in NDI scores compared with baseline (MD in change scores −0.97, 95% CI −7.15 to 5.21, scale unclear), SF-36 PCS scores (MD in change scores −1.90, 95% CI −5.30 to 1.50, 0-100 scale) and SF-36 MCS scores (MD in change scores 0.42, 95% CI −2.30 to 3.14, 0-100 scale).⁴⁴ One retrospective NRSI (N=245) compared ACDF (with and without corpectomy) with PCDF and reported median NDI scores (16 vs. 17, adjusted OR 0.76, 95% CI 0.42 to 1.37)⁴¹ (SOE: Low).

3.7.3.6.1. Neurologic Deficits

There was low-strength evidence that posterior approaches were more likely associated with a moderate to large increase in the odds of experiencing a neurologic adverse event compared with ACDF (SOE: Low). Reporting of neurological events varied across one RCT (N=32)⁴⁰ and six NRSIs (total N=37,095, range 245 to 13,884).^{41–44,48,49} The RCT reported no cases of postoperative worsening of myelopathy or C5 root palsy with either 3- or 4-level ACDF versus posterior laminoplasty.⁴⁰ Central nervous system complications (not further defined) were rare through 90 days after ACDF (<0.7%) and posterior laminoplasty (0.9%) at three or more levels in one NRSI (N=3,042).⁴⁹ Two NRSIs reported that PCDF was associated with moderately higher odds of “neurological complications” compared with ACDF at three or more levels but did not provide further details: 0.59% vs. 0.35% (adjusted OR 1.7, 95% CI 1.0 to 2.8) immediately postoperative in one study (N=13,884)⁴² and 1.8% vs. 1.1% (OR 1.6, 95% CI 1.08 to 2.38) at 1 month in another (N=7,412).⁴³ Two other NRSIs reported no difference between ACDF and PCDF at three to five levels in new neurological deficits (N=264, 4.1% vs. 3.2%, RR 1.31, 95% CI 0.35 to 4.95)⁴⁴ or new motor deficits (N=245, 2% vs. 0%)⁴¹ at 12 months. One large NRSI (N=12,248) reported no difference between PCDF and ACDF in the incidence of postoperative coma (0.4% vs. 0.6%, OR 1.26, 95% CI 0.75 to 1.77).⁴⁸

3.7.3.6.2. Mortality

There was low-strength evidence that mortality did not differ between ACDF and laminoplasty or PCDF (SOE: Low).

Three NRSIs (total N=15,057, range 546 to 13,884) that compared anterior with posterior approaches at three or more levels found no difference in short-term mortality after ACDF versus posterior laminoplasty at 1 month (1 NRSI, N=364, 0% vs. 0.05%, RR 0.33, 95% CI 0.01 to 8.13)⁴⁷ and ACDF versus PCDF at hospital discharge to 1 month (3 NRSIs, N=14,875, 0.3% vs. 0.3%, RR 0.96, 95% CI 0.25 to 1.81, I²=17.8%)^42,46,47 (Figure 6). One NRSI (N=12,248) reported no deaths in either arm (ACDF vs. PCDF) and was unable to be included in the pooled analysis.⁴⁸

Figure 6

Mortality: anterior versus posterior approaches for ≥3 levels. ACDF = anterior cervical discectomy and fusion; CI = confidence interval; CSM = cervical spondylotic myelopathy; OPLL = ossification of the posterior longitudinal ligament, PCDF = (more...)

3.7.3.6.3. Dysphagia

There was low-strength evidence that the likelihood of experiencing severe dysphagia did not differ between ACDF and laminoplasty or PCDF (SOE: Low).

Severe dysphagia was rare across two NRSIs that compared ACDF with PCDF or posterior laminoplasty. There were two cases (1%) requiring a nasogastric tube in one study (N=245)⁴¹ and one case (0.5%) requiring an unplanned readmission 11 days post surgery in the other (N=364);⁴⁷ all three cases occurred in the ACDF arms (SOE: Low).

One RCT (N=32)⁴⁰ and seven NRSIs (total N=41,172, range 245 to 13,884)^{41–43,45,46,48,49} also reported dysphagia but did not report the severity; frequencies ranged from 2.7 to 14.0 percent after ACDF and from 0 to 3.6 percent after PCDF across six NRSIs (N=38,130),^{41–43,45,46,48} most of which reported a substantial to moderate decrease in the odds/risk of dysphagia with PCDF (OR range 0.20 to 0.61), and from <0.7 to 5.9 percent versus 0 to <0.7 percent in the ACDF versus laminoplasty arms, respectively, across one small RCT (N=32)⁴⁰ and one large NRSI (N=3,042), with no differences between treatments.⁴⁹

3.7.3.6.4. Serious Adverse Events

There was low-strength evidence that posterior approaches were more likely associated with a moderate to large increase in the odds of experiencing a serious adverse event compared with ACDF (SOE: Low).

One RCT (N=32) reported that intraoperative dural tear occurred in 5.9 percent of ACDF versus 11.8 percent of PCDF patients (RR 0.50, 95% CI 0.05 to 5.01) and that there were no cases of instrumentation failure or malposition, infection or hematoma.⁴⁰

Across the NRSIs, reporting of serious adverse events varied; adverse events generally occurred more often with posterior approaches versus ACDF.

Thrombolic events were rare across eight NRSIs (total N=41,718, range 245 to 13,884) with followup immediately postoperative to 12 months.^{41–43,45–49} The frequency of deep vein thrombosis (DVT) or pulmonary embolism ranged from 0 to 2.3 percent (ACDF) versus 0 to 4.3 percent (PCDF or posterior laminoplasty). Four of the studies (N=37,258) reported that posterior approaches were associated with moderate to large increases in the odds of experiencing a thrombolic event compared with ACDF (range of ORs 1.75 to 3.7).^42,43,45,48

Stroke/cerebrovascular events occurred variably across three NRSIs with short-term followup (1 to 3 months); one study (N=546) reported no events in either arm (ACDF vs. PCDF or posterior laminoplasty),⁴⁷ one study (N=627) reported more events after ACDF (1.8% vs. 0% PCDF, p=0.016),⁴⁶ while the third found that PCDF was associated with a large increase in the odds of stroke compared with ACDF (N=12,248, 4.2% vs. 2.5%, OR 1.68, 95% CI 1.48 to 1.89).⁴⁸

Sepsis was rare across three NRSIs (total N=7,302, range 546 to 3,714).^45,47,49 One study reported substantially higher odds of having sepsis within 3 months after PCDF compared with ACDF (N=3,714, 2.5% vs. 0.7%, adjusted OR 3.56, 95% CI 1.96 to 6.91)⁴⁵ while the other two studies (N=3,588) reported similar rates between groups (ACDF, range <0.7% to 1.1% vs. PCDF/posterior laminoplasty, range <0.7% to 1.7%)^47,49

Surgical site infection was reported by four NRSIs. Three studies (N=22,702)^43,48,49 reported that posterior approaches (PCDF or laminoplasty) were associated with a large increase in the odds of surgical site infection compared with ACDF at 1 to 3 months (frequency range 2.4% to 4.7% vs. 0.8% to 1.0%, OR range 3.1 to 3.7) and the fourth (N=245) found no difference between groups (1% each).⁴¹

Wound dehiscence was infrequent across four NRSIs, two of which reported that PCDF was associated with a substantial increase in the odds of experiencing this complication compared with ACDF (N=19,660, frequency range 1.3% to 2.7% vs. 0.1% to 0.5%, range of ORs 5.6 to 10.8)^43,48 and two that found no difference between groups (1% each, N=245, 1 RCT)⁴¹ and (0% each, N=264, 1 RCT).⁴⁴

Dural tear/durotomy occurred more often with ACDF versus PCDF in one study (N=627, 9.4% vs. 3.2%, RR 3.02, 95% CI 1.50 to 6.10)⁴⁶ while no events were reported in either group in another study (N=264).⁴⁴

One NRSI found that PCDF was associated with a large increase in the odds of having any severe adverse event through 3 months compared with ACDF (N=3,714, 13% vs. 6.1%, OR 2.31, 95% CI 1.83 to 2.93).⁴⁵

A variety of other serious adverse events were reported across five NRSIs (total N=21,813, range 546 to 13,884);^{42,45–47,49} event rates ranged from 0.04 to 4.5 percent in the ACDF arms and from 0 to 7.7 percent in the posterior arms (PCDF or laminoplasty) and included kidney injury (4 studies)^45–47,49 cardiac complications (4 studies),^42,46,47,49 transfusion (3 studies),^45–47 respiratory complications (3 studies),^42,46,49 and arterial injury and hardware instrument failure malposition (1 study).⁴² Excluding perioperative blood transfusion in one study, which had the highest frequency of events across all these complications (N=627, 4.5% with ACDF vs. 7.7% with a posterior approach),⁴⁶ the range across treatment arms was 0 to 3.7 percent (ACDF) versus 0.06 to 3.6 percent (posterior approach). There were no cases of myocardial infarction or vocal cord paralysis in one NRSI (N=245).⁴¹

3.8.3.3.1. Neurologic Function

There was moderate-strength evidence of no difference between laminectomy and fusion versus laminoplasty on neurologic function (SOE: Moderate).

Two head-to-head RCTs (N=46) assessed neurologic function with the mJOA and the Nurick Classification Scale for Spinal Cord Compression (i.e., Nurick’s grade 0 to 5) at 1 year post-operative.^50,51 Pooled analysis of the two trials found no difference in function between cervical laminectomy and fusion versus laminoplasty using the mJOA (N=46, MD −0.03, 95% CI −0.68 to 0.74, I²=76%).^50,51 One trial reported no significant difference between laminectomy and fusion compared with laminoplasty in Nurick grade (1.40 vs. 1.67; p=0.23),⁵⁰ while the other trial reported a significant pre-post difference for laminoplasty only (numeric values not reported; p<0.05).⁵¹

Four nonrandomized studies reported neurologic function using the mJOA or JOA score; three reported no difference between laminectomy and fusion versus laminoplasty^52,54,57 and one reported a significant benefit of laminoplasty over laminectomy and fusion (mean mJOA at 2 years: 3.49, 95% CI 2.84 to 4.13 vs. 2.39, 95% CI 1.91 to 2.86; p=0.0069).⁵³ However, this study reported no significant difference in Nurick’s grade at 2 years (mean 1.57, 95% CI 1.23 to 1.90 vs. 1.18, 95% CI 0.92 to 1.44; p=0.077).

3.8.3.3.2. General Function

There was low-strength evidence of little difference between laminectomy and fusion versus laminoplasty on general function (SOE: Low).

Neck disability scores on the NDI were not different between laminectomy and fusion versus laminoplasty 1-year postoperatively (1 RCT, N=30, MD 3.86, p=0.2)⁵⁰ and only improved with laminoplasty in the other trial (N=16, surgical approaches not directly compared, numeric values not reported, p=0.05).⁵¹ The same trial (N=16) reported improvement from baseline on the SF-36 with laminoplasty only (numeric values not reported, p<0.05).^51,52,54 Two NRSIs reported no differences on the NDI,^52,53 and three reported no differences between surgical approaches in SF-12 or SF-36 PCS or MCS scores.^52–54 Another observational study reported no differences in improved gait (71% vs. 68%; p=0.674) as assessed on a 5-point NRS.⁵⁷

3.9.3.1.1. Fusion

Seven RCTs (across 15 publications) (N=2,382) that compared single-level cervical arthroplasty and ACDF reported fusion success in their ACDF arms.^{59,60,68,75–78,86,87,92–95,98,101} One trial (N=56) reported short-term fusion success in 89.3 percent of participants,¹⁰¹ seven RCTs (N=853) reported intermediate-term fusion success in 93.9 percent (range 89.1% to 98.2%) of participants^{59,68,75,78,92,98,101} and two RCTs (N=181) reported long-term fusion success in 96.5 percent (range 95.5% to 96.9%) of participants.^60,95 One RCT reported successful fusion in the cervical arthroplasty arm as well, but this may be attributed to participant crossover after initial randomization.^92,93

3.9.3.1.2. Pain

3.9.3.1.2.1. Neck Pain

There was moderate-strength evidence of no differences between cervical arthroplasty and ACDF in neck pain or likelihood of success (response) for neck pain at short, intermediate, and long-term (SOE: Moderate).

Four RCTs (N=1,230) (in 5 publications)^{92,97,114,118,119} that compared single level cervical arthroplasty versus ACDF reported neck pain success (response) defined as postoperative ≥20-point improvement on VAS. There were no differences in likelihood of neck pain success between cervical arthroplasty and ACDF at short term (2 RCTs, N=482, 79% vs. 75.0%, RR 1.04, 95% CI 0.93 to 1.17, I²=0%),^114,119 intermediate term (4 RCTs, N=948, 76.4% vs. 74.1%, RR 1.03, 95% CI 0.95 to 1.12, I²=0%)^{92,114,118,119} or long term (1 RCT, N=232, 85.7% vs. 78.3%, 1.09, 95% CI 0.97 to 1.24)⁹⁷ (Figure 7). In one prospective NRSI IDE study using propensity-matched historical controls, more cervical arthroplasty participants had≥20-point improvement on VAS neck pain versus ACDF at 24 months (N=301, 91.2% vs. 77.9%, p=0.013).¹²⁰

One of the above trials reported neck pain success at 84 months using an alternative definition, a ≥10-point improvement on VAS, and was not included in the meta-analysis at long term; there was no difference between cervical arthroplasty and ACDF using this criterion (N=191, 87.5% vs. 83.3%, RR 1.05, 95% CI 0.93 to 1.20).⁹⁵

Figure 7

Neck pain success (≥20-point improvement on VAS): comparison of cervical arthroplasty with ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = (more...)

Eleven RCTs (N=2,696) (in 19 publications)^{60,61,67,69,75,78,79,81,84,86,88,89,92,95–98,100,116} contributed to evaluation of mean differences in neck pain scores at various times. There were no differences between cervical arthroplasty and ACDF in VAS neck pain scores (0-100 scale) as estimates were below the threshold for a small effect at short term (8 RCTs, N=1,789, MD −3.02, 95% CI −5.53 to 0.40, I²=15.5%),^{61,69,75,78,86,89,98,116} intermediate term (11 RCTs, N=1,898, MD −3.39, 95% CI −6.14 to −1.23, I²=63.4%),^{60,61,67,69,78,79,88,92,96,98,100} and long term (5 RCTs, N=1,195, MD −4.77, 95% CI −7.62 to −1.72, I²=0%)^{60,81,84,95,97} (Figure 8). Exclusion of one, small (N=60) trial rated high risk of bias⁶¹ did not substantially change effect estimates but did slightly increase heterogeneity in the short term (7 RCTs, N=1,729, MD −3.11, 95 % CI −5.92 to −0.15, I²=26.6%)^{69,75,78,86,89,98,116} and intermediate term (10 RCTs, N=1,838, MD −3.55, 95% CI −6.48 to −1.30, I²=67.1%).^{60,67,69,78,79,88,92,96,98,100} Exclusion of one trial⁶⁹ that did not specify if neck or arm pain was evaluated also did not substantially change effect estimates at short term (7 RCTs, N=1,714, MD −3.24, 95% CI −5.95 to −0.77, I²=12.2%)^{61,75,78,86,89,98,116} or intermediate term (10 RCTs, N=1,879, MD −3.51, 95% CI −6.35 to −1.33, I²=66.4%).^{60,61,67,78,79,88,92,96,98,100} Although funnel plot analysis and Egger’s test (p=0.035) may suggest publication/small study bias for neck pain scores at intermediate term, most trials found no effect leading to less concern regarding publication bias (Appendix F, Figure F-4).

Figure 8

Neck pain VAS scores (0-100 scale): comparison of cervical arthroplasty with ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; (more...)

3.9.3.1.2.2. Arm Pain

There was moderate-strength evidence of no differences between cervical arthroplasty and ACDF in arm pain or likelihood of success (response) for arm pain at short, intermediate, and long-term (SOE: Moderate).

Four RCTs (N=1,148) (in 5 publications)^{92,97,114,118,119} that compared cervical arthroplasty with ACDF for single level disease reported arm pain success (response) defined as postoperative ≥20-point improvement on VAS (0–100). Some studies reported arm pain success in both arms. Conservative estimates, using the lower risk ratio for studies reporting VAS for both arms, revealed no difference in likelihood of arm pain success between cervical arthroplasty and ACDF at short term (2 RCTs, N=482, 49.5% vs. 46.6%, RR 1.02, 95% CI 0.81 to 1.29, I²=0%),^114,119 intermediate term (4 RCTs, N=948, 61.1% vs. 62.6%, RR 1.0, 95% CI 0.85 to 1.14, I²=37.9%),^{92,114,118,119} or long term (1 RCT, N=232, 85.7% vs. 75.5%, RR 1.14, 95% CI 1.00 to 1.29, I²=0%)⁹⁷ (Figure 9). Estimates based on higher risk ratios for studies reporting VAS for both arms were similar and led to the same conclusion of no difference between cervical arthroplasty and ACDF for all time points. In one prospective NRSI IDE study using propensity-matched historical controls, more cervical arthroplasty participants experience≥20-point improvement on VAS arm pain (worst side) versus ACDF at 24 months (N=301, 90.5% vs. 79.9%, p=0.001).¹²⁰

Figure 9

Arm pain success (≥20-point improvement on VAS): comparison of cervical arthroplasty with ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = (more...)

Nine RCTs (N=2,460) (in 17 publications)^{60,67,75,78,81,84,86,88–90,92,95–98,100,116} assessed arm pain at various times. Three publications reported pain scores for both arms. Using a conservative estimate with the smaller effect estimate of the two arms, there was no difference between cervical arthroplasty and ACDF in VAS arm pain scores (0-100 scale) short term (6 RCTs, N=1,761, MD −0.66, 95% CI −2.93 to 1.43, I²=0%),^{75,78,86,89,98,116} intermediate term (9 RCTs, N=1,741, MD −1.86, 95% CI −4.03 to −0.60, I²=0%),^{60,67,78,88,90,92,96,98,100} or long term (5 RCTs, N=1,195, MD −4.55, 95% CI −7.62 to −1.68, I²=0%)^{60,81,84,95,97} (Figure 10). Exclusion of one small (N=20) trial rated high risk of bias⁹⁰ did not impact the effect size. Using the larger effect estimate when both arms were measured, slightly increased the estimate at short term but not the conclusion of no difference between treatments (MD −1.11, 95% CI −3.56 to 1.02); estimates at intermediate and long term were similar to the conservative estimates.

Figure 10

Arm pain VAS scores (0-100): comparison of cervical arthroplasty with ACDF (1-level). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; FDA = Food and Drug Administration; (more...)

3.9.3.1.3. Function

3.9.3.1.3.1. Neurologic Function

There was moderate-strength evidence of no differences between cervical arthroplasty and ACDF in neurologic function at short, intermediate, and long term (SOE: Moderate).

Six RCTs (N=2,271) (in 15 publications)^{60,78,81,84,86,92,95–98,102,114,116,118,119} that compared single-level cervical arthroplasty and ACDF reported neurologic success (response) defined as maintenance or improvement (compared with preoperative status) in all three of the following areas: motor function, sensory function and deep tendon reflexes. There were no differences between cervical arthroplasty and ACDF in the likelihood of neurological success short-term (5 RCTs, N=1,493, 95.2% vs. 90.5%, RR 1.04, 95% CI 1.01 to 1.08, I²=0%),^{86,114,116,118,119} intermediate term (6 RCTs, N=1,574, 93.3% vs. 89.5%, RR 1.03, 95% CI 1.00 to 1.06, I²=0%),^{60,78,92,96,98,102} or long term (5 RCTs, N=1,180, 89.9% vs. 86.6%, RR 1.02, 95% CI 0.97 to 1.09, I²=43.3%)^{60,81,84,95,97} (Figure 11). One prospective NRSI IDE study that used propensity matched ACDF historical controls reported neurological success, defined as maintenance or improvement compared with baseline, was similar for cervical arthroplasty and ACDF at 24 months (N=314, 99.3% vs. 98.8%).¹²⁰

Figure 11

Neurological success: comparison of cervical arthroplasty with ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; FDA = Food and (more...)

Four RCTs (N=354), three rated high risk of bias^63,91,101 and one low risk of bias,⁷⁹ reported JOA scores (0-17). There was no differences between cervical arthroplasty and ACDF in pooled analysis at intermediate term (4 RCTs, N=354, MD 0.60, 95% CI −0.007 to 0.97, I²=1.9%) or in one short-term trial rated high risk of bias (1 RCT, N=60, MD 0.25, 95% CI −0.25 to 0.75).⁶³

One trial reported the proportion of participants who had the same or an improved Nurick grade at 60 months compared with baseline; there were no differences (i.e., point estimate below the threshold for a small effect) between cervical arthroplasty and ACDF (N=285, 99.4% vs. 96.9%, RR 1.03, 95% CI 0.99 to 1.06).⁹³

3.9.3.1.3.2. General Function

There was moderate-strength evidence of no differences between cervical arthroplasty and ACDF in general function at short, intermediate, and long term (SOE: Moderate).

3.9.3.1.3.2.1. NDI

Six RCTs (N=2,271) (in 14 publications)^{60,78,84,86,87,92,95–98,114,116,118,119} that compared cervical arthroplasty with ACDF for single-level disease reported NDI success (response) defined as postoperative NDI score improvement of ≥15 points from the baseline score (FDA definition). There were no differences between cervical arthroplasty and ACDF in the likelihood of NDI success short term (6 RCTs, N=1,900, 85.2% vs. 79.0%, RR 1.07, 95% CI 1.01 to 1.13, I²=31.6%),^{86,96,114,116,118,119} intermediate term (6 RCTs, N=1,678, 82.9% vs. 78.2%, RR 1.07, 95% CI 1.01 to 1.14, I²=8.4%),^{60,78,87,92,96,98} or long term (4 RCTs, N=1,047, 86.4% vs. 80.8%, RR 1.06, 95% CI 0.99 to 1.15, I²=35.5%)^60,84,95,97 (Figure 12). In one prospective NRSI IDE study that used propensity-matched historical controls, there was no difference in NDI success (≥15-point NDI improvement) following cervical arthroplasty versus ACDF at 24 months (N=301, 90.5% vs. 85.1%, p=0.372).¹²⁰

Figure 12

NDI success (≥15-point improvement): comparison of cervical arthroplasty with ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; (more...)

Twelve RCTs (N=2,800) (in 19 publications)^{60,61,67,69,75,78,79,81,82,84,86,92,93,95–98,100,101} that compared cervical arthroplasty with ACDF reported NDI scores (0-100 scale). There were no differences between cervical arthroplasty and ACDF in NDI scores as estimates were below the threshold for a small effect at short term (8 RCTs, N=2,125, MD −3.13, 95% CI −4.29 to −1.99, I²=0%),^{61,67,69,75,78,86,93,97} intermediate term (12 RCTs, N=2,027, MD −2.10, 95% CI −3.94 to −0.35, I²=49.3%),^{60,61,67,69,78,79,82,92,96,98,100,101} or long term (6 RCTs, N=1,291, MD −3.30, 95% CI −5.13 to −1.02, I²=0%)^{60,69,81,84,95,97} (Figure 13). Exclusion of trials rated high risk of bias^61,82,101 had no impact on effect estimates or statistical heterogeneity in the short term (7 RCTs, N=2,065, MD −3.14, 95% CI −4.30 to −1.99, I²=0%)^{67,69,75,78,86,93,97} and slightly increased effect size and increased heterogeneity at intermediate term (9 RCTs, N=1,814, MD −2.45, 95% CI −4.70 to −0.35, I²=62.5%).^{60,67,69,78,79,92,96,98,100} Exclusion of a trial rated moderate risk of bias⁶⁹ with unclear sample sizes resulted in a small increase in effect size long term (5 RCT, N=1,288, MD −3.78, 95% CI −5.74 to −1.54).^{60,81,84,95,97} There was no indication of publication/small study bias for NDI scores at intermediate term based on funnel plot analysis (Egger’s test, p=0.416) (Appendix F, Figure F-5).

Figure 13

NDI scores (0-100): comparison of cervical arthroplasty with ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; FDA = Food and Drug (more...)

3.9.3.1.3.2.2. SF-36 and SF-12 PCS and MCS

Four RCTs (N=1,148) (in 6 publications)^{92,97,98,114,118,119} that compared cervical arthroplasty with ACDF for single-level disease reported SF-36 and SF-12 PCS and MCS (0-100 scale). Success for these component scores was defined as postoperative score improvement of ≥15 points from baseline scores. The likelihood of PCS success was similar for cervical arthroplasty and ACDF short term (2 RCTs, N=466, 81.7% vs. 75.9%, RR 1.08, 95% CI 0.96 to 1.23, I²=0%),^114,119 intermediate term (4 RCTs, N=939, RR 1.16, 95% CI 1.00 to 1.41, I²=61.2%),^{92,98,114,118} and long term (1 RCT, N=231, 72.0% vs. 74.5%, 0.97, 95% CI 0.83 to 1.13)⁹⁷ (Figure 14). Exclusion of one outlier trial¹¹⁸ at intermediate term resulted in a slightly attenuated effect estimate but did not reduce heterogeneity or change the conclusion (3 RCTs, N=750, RR 1.12, 95% CI 0.96 to 1.34, I²=59.8%).^92,98,114 In one prospective NRSI IDE study using propensity-matched historical controls, more cervical arthroplasty participants maintained or improved PCS score versus ACDF at 24 months (N=301, 97.3% vs. 89.2%, p=0.023).¹²⁰ The likelihood of MCS success was also similar for cervical arthroplasty and ACDF at all time points: short term (2 RCTs, N=466, 49.1% vs. 42.8%, RR 1.13, 95% CI 0.86 to 1.50, I²=0%),^114,119 intermediate term (4 RCTs, N=939, 47.3% vs. 48%, RR 0.97, 95% CI 0.80 to 1.16, I²=27.5%)^{92,98,114,118} and long term (1 RCT, N=231, 47.2% vs. 43.4%, RR 1.09, 95% CI 0.82 to 1.45)⁹⁷ (Figure 15). In the prospective NRSI IDE study, there was no difference in MCS maintenance or improvement between procedures at 24 months (N=301, 77.6% vs. 77.0%).¹²⁰

Figure 14

SF-36 or SF-12 PCS success (≥15-point improvement): comparison of cervical arthroplasty with ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI (more...)

Figure 15

SF-36 or SF-12 MCS success (≥15-point improvement): comparison of cervical arthroplasty with ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI (more...)

Seven RCTs (N=2,368) (in 14 publications)^{60,69,75,77,78,81,84,86,92,95–98,116} that compared cervical arthroplasty with ACDF reported SF-36/12 PCS and MCS scores (0-100 scale). There were no differences between cervical arthroplasty and ACDF in PCS scores (Figure 16) as estimates were below the threshold for a small effect in the short-term (6 RCTs, N=1,779, MD 1.67, 95% CI 0.59 to 2.87, I²=0%), intermediate term (7 RCTs, N=1,684, MD 2.13, 95% CI 0.77 to 3.33, I²=0%), or long term (5 RCTs, N=1,191, MD 1.76, 95% CI 0.44 to 3.07, I²=0%). Similarly, there were no differences between cervical arthroplasty and ACDF in MCS scores (Figure 17) as estimates were below the threshold for a small effect in the short-term (6 RCTs, N=1,779, MD 1.14, 95% CI −0.14 to 2.17, I²=0%), intermediate term (7 RCTs, N=1,814, MD 0.83, 95% CI −0.75 to 2.41, I²=32.2%), and long term (3 RCTs, N=574, MD 0.64, 95% CI −1.47 to 2.82, I²=0%). Effect estimates for PCS and MCS did not differ following the exclusion of one trial with unclear samples sizes.⁶⁹ No studies were rated high risk of bias.

Figure 16

SF-36 or SF-12 PCS scores (0-100): comparison of cervical arthroplasty with ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; FDA (more...)

Figure 17

SF-36 or SF-12 MCS scores (0-100): comparison of cervical arthroplasty with ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; FDA (more...)

3.9.3.1.3.2.3. Odom’s Criteria

Four RCTs (N=553)^61,82,91,93 used Odom’s criteria to categorize overall improvement as excellent (i.e., all pre-operative symptoms relieved, abnormal findings improved), good (i.e., minimal persistence of symptoms, abnormal findings unchanged or improved), fair (i.e., definite relief of some symptoms, others unchanged or slightly improved) or poor (i.e., symptoms and signs unchanged or exacerbated). There were no differences between single-level cervical arthroplasty and ACDF in the likelihood of having excellent or good results based on Odom’s criteria (4 RCTs, N=847, 48.3% vs. 46.8%, RR 1.01, 95% CI 0.92 to 1.12, I²=0%) at intermediate term.^61,82,91,93 However, three of the RCTs (all small) were rated high risk of bias,^61,82,91 while the one large RCT was rated moderate risk of bias.⁹³ Based on the highest quality trial, there was no difference between procedures in the likelihood of having excellent or good improvement (1 RCT, N=682, 45.7% vs. 43.1%)⁹³ (Figure 18). In one prospective NRSI IDE study using propensity-matched historical controls, there was no difference between cervical arthroplasty and ACDF in the likelihood of having excellent or good results using Odom’s criteria at 24 months (N=301, 90.5% vs. 79.9%).¹²⁰

Figure 18

Odom’s criteria: comparison of cervical arthroplasty with ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; mos. = months; (more...)

3.9.3.1.3.3. Overall Success (Composite)

The FDA IDE trials were required to report overall success, a composite outcome for six RCTs (N=2,271) (in 11 publications)^{60,75,78,84,86,87,93,96,98,118,119} that included a threshold of ≥15-point NDI improvement (0-50 scale) from baseline, improvement or maintenance of neurologic status, no serious adverse events and no additional surgical procedures that might be considered “failure” (e.g., removal, revision, supplemental fixation). In participants with single-level interventions, effect estimates were below the threshold for a small effect and classified as no difference in overall success comparing cervical arthroplasty with ACDF in the short term (4 RCTs, N=1,361, 79.9% vs. 71.7%, RR 1.11, 95% CI 1.04 to 1.18, I²=0%)^{75,86,118,119} and intermediate term (6 RCTs, N=1,717, 76.1% vs. 67.7%, RR 1.14, 95% CI 1.07 to 1.20, I²=0%);^{60,78,87,93,96,98} but a slightly increased likelihood of overall success favoring cervical arthroplasty was seen long term (3 RCTs, N=878, 76.1% vs. 67.7%, RR 1.21, 95% CI 1.11 to 1.32, I²=0%)^60,84,98 (Figure 19). In one prospective NRSI IDE study using propensity-matched historical controls, there was no difference between cervical arthroplasty and ACDF in overall response (same definition as in RCTs) at 24 months (N=301, 86.8% vs. 79.3%, p=0.265).¹²⁰

One of the above trials reported overall success at 84 months using a different criterion for NDI (improvement in NDI score ≥30 points if preoperative score ≥60 or improvement of ≥50% if preoperative score <60) and included an additional requirement for radiographic success, and was not included in the meta-analysis at long term; there was no difference between cervical arthroplasty and ACDF using this criteria (N=166, 55.2% vs. 50.0%, RR 1.10, 95% CI 0.80 to 1.52).⁹⁵

Figure 19

Overall success: comparison of cervical arthroplasty with ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; FDA = Food and Drug (more...)

3.9.3.1.3.5. Reoperation and Subsequent Surgery

There was high-strength evidence that cervical arthroplasty was associated with substantially lower likelihood of reoperation that included the index level versus ACDF (SOE: High). Rates of reoperation for ACDF at the index level may be influenced by the need to remove an existing plate to treat adjacent segment disease (ASD), rather than the indication for reoperation being driven by an issue at the index procedure. This may artificially inflate the reported reoperation rate at the index procedure level for ACDF when compared with cervical arthroplasty. Studies were not consistently clear in the indication for reoperation. The clinical relevance of removing the plate as a part of a procedure addressing ASD is minimal.

Reoperation including any additional procedure at the index level was substantially less frequent with cervical arthroplasty versus ACDF for single-level disease at all time points reported in RCTs including short term up to 24 months (9 RCTs, N=2,323, 2.9% vs. 6.2%, RR 0.49, 95% CI 0.28 to 0.80, I²=16.2%)^{60,76,82,87,90,96,98,100,116} and long term from 84 to 120 months (7 RCTs, N=1,992, 5.2% vs. 12.5%, RR 0.44, 95% CI 0.29 to 0.60, I²=0%)^{60,69,81,85,92,95,97} (Figure 20).

Figure 20

Reoperation involving the index level: comparison of cervical arthroplasty with ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; (more...)

One prospective NRSI IDE study of cervical arthroplasty using historical ACDF controls found no difference in index-level reoperation up to 24 months (N=349, 1.9% vs. 4.8%, RR 0.39, 95% CI 0.11 to 1.43).¹²⁰

Reoperation across two NRSIs was less common than that reported in RCTs. No difference in 30-day reoperation was seen in one NRSI (1.2% vs. 0.4%, adjusted OR 0.60, 95% CI 0.14 to 2.56).¹⁰⁹ Another NRSI reported that reoperation was less common following cervical arthroplasty within 90 days of index surgery compared with ACDF (2.04% vs. 3.35%, adjusted OR 0.63, 95% CI 0.44 to 0.92) but no difference between cervical arthroplasty and ACDF longer-term up to 5 years (adjusted hazard ratio 0.86, 95% CI 0.60 to 1.23).¹⁰⁸ While overall reoperation rates were lower in these database NRSIs, it is possible the RCTs, particularly IDE trials may provide more accurate detail regarding specific indications.

Subsequent surgery rates at adjacent levels were similar between cervical arthroplasty and ACDF at up to 24 months^{82,86,98,100,114–116,118} and between 36 and 48 months (including after exclusion of one trial rated high risk of bias¹⁰¹)^{89,96,101,116} but was substantially less likely with cervical arthroplasty versus ACDF at 60 months (3 RCTs, N=1,010, 2.5% vs. 6.2%, RR 0.39, 95% CI 0.15 to 0.84, I²=8.7%)^59,68,80 and at the longest followups from 84 to 120 months (6 RCTs, N=1,606, 5.0% vs. 13.5%, RR 0.39, 95% CI 0.25 to 0.56, I²=1.5%).^{60,69,81,85,95,97} However, estimates were somewhat imprecise (Figure 21). Also, across trials, indications for operation at adjacent levels were not consistently described.

Figure 21

Subsequent surgery at adjacent levels: comparison of cervical arthroplasty versus ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; (more...)

3.9.3.1.3.6. Harms

All 15 RCTs that evaluated cervical arthroplasty and ACDF for single-level disease provided information on adverse events and harms up to 120 months followup.^{60,61,69,76,79,82,87,89,90,92,96,98,100,101,116} Information on harms from four NRSIs was used to complement that from RCTs.^{106,108,109,120}

3.9.3.1.3.6.1. Neurologic Deficit

There was low-strength evidence of no differences in the likelihood of neurological events or deficits between cervical arthroplasty and ACDF in the short, intermediate, or long term (SOE: Low).

Reporting of neurological events varied across RCT publications. Three publications assessed events from the Bryan IDE trial at different times;^58,85,96 one IDE trial evaluated Mobi-C.⁹⁵ One trial⁵⁸ described specific, observed neurological events as acute neurological changes, while other trials used various general terms to describe neurologic events (e.g., new deficit, neurological failure, neurological adverse event). The timing of events following surgery was also not clearly reported. Thus, reported proportions of participants who experienced neurological events varied substantially across RCTs, however there were no differences between cervical arthroplasty and ACDF at 0 to 24 months (3.3% vs. 3.2%),⁵⁸ between 24 and 48 months (0% vs. 1.0%, WHO grade 3 or 4),⁹⁶ up to 84 months (11.4 % vs. 11.5%),⁹⁵ or up to 120 months (any: 43.1% vs. 43.8%; WHO grade 3 or 4: 4.5% vs. 6.9%).⁸⁵ One prospective NRSI IDE study of cervical arthroplasty that used propensity-matched historical ACDF controls reported no differences in serious device- or procedure-related neurological adverse events between cervical arthroplasty and ACDF (1.3% vs. 1.6%) through 24 months.¹²⁰ The same trial study also reported fewer cervical arthroplasty participants experienced neurological decrease from baseline versus ACDF (6.7% vs. 12.8%, RR 0.52, 95% CI 0.25 to 1.07) but results were imprecise.

3.9.3.1.3.6.2. Mortality

There was inadequate evidence to draw conclusions on the likelihood of death in participants undergoing cervical arthroplasty versus ACDF (SOE: Insufficient).

Death was uncommon (<3%) in RCTs and NRSIs, with no reported differences between cervical arthroplasty and ACDF. Across RCTs, no deaths were directly attributed to either procedure, however cause of death was not reported in many trials. For cervical arthroplasty from 0 to 24 months, three of the four deaths were attributed to myocardial infarction or cardiac arrest in one trial;⁶⁰ the cause of the fourth death was not reported in another trial.⁹⁸ No deaths were observed in one trial.⁷⁶ At followup from 0 to 36 months, one cervical arthroplasty participant died of a severe subarachnoid hemorrhage at 6 weeks (relationship to procedures was not stated)⁸⁹ and one death in the ACDF group attributed to a motor vehicle accident was observed in another trial.⁵⁸ There was no difference in mortality between procedures at 84 months (1 RCT, N=541, 0.9% vs. 2.2%, RR 0.38, 95% CI 0.08 to 1.96)⁶⁰ or at 120 months (1 RCT, N=232, 1.4% vs. 2.4%, RR 0.54, 95% CI 0.09 to 3.18),⁸⁵ however estimates were imprecise. Findings from one large administrative data NRSI¹⁰⁸ reinforce that death was rare for cervical arthroplasty (0%) and ACDF (0.18%) and that there was no difference between procedures in the likelihood of mortality. One death occurred in the cervical arthroplasty group in one NRSI IDE study using historical controls up to 24 months¹²⁰ (Appendix C).

3.9.3.1.3.6.3. Serious Adverse Events

There was low-strength evidence that cervical arthroplasty was associated with a slightly lower likelihood of any serious adverse event in the short term versus ACDF (SOE: Low); there was also low-strength of no differences in the likelihood of experiencing a serious adverse events at greater than 24 months (SOE: Low).

Serious adverse event definitions and types of events varied across RCTs, but often included events that were life threatening, required medical intervention, or resulted in a permanent disability or death. Timing of events was not reported. Events related to participant factors such as comorbidities (e.g., underlying cardiovascular disease) would likely not be different between procedures. Cervical arthroplasty was associated with a slightly lower likelihood of experiencing a serious adverse event up to 24 months across IDE trials (5 RCTs, N=1,611, 24.6% vs. 30.6%, RR 0.83, 95% CI 0.64 to 0.97, I²= 24.6%)^{58,76,87,93,98} compared with ACDF, however across fewer trials at other times, no differences between procedures was seen (Figure 22). No difference in the likelihood of experiencing a serious adverse events was seen between cervical arthroplasty and ACDF (N=349, 9.4% vs. 14.8%, RR 1.97, 95% CI 0.88 to 4.37) in one NRSI IDE study using historical controls up to 24 months.

Figure 22

Any serious adverse events (author defined): comparison of cervical arthroplasty with ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; (more...)

Dysphagia was reported by six RCTs (N=1,965) (in 8 publications),^{58,60,68,69,76,81,85,98} but the severity was unclear in most cases. One trial (N=463) reported no cases of WHO grade 3 or 4 dysphagia in any participant through 24 months followup.⁵⁸

NRSIs based on administrative data suggest that serious adverse events are rare and not different between cervical arthroplasty and ACDF. Thrombolic event rates (DVT and/or pulmonary embolism) were similar between cervical arthroplasty (range 0.07% to 0.19%) and ACDF (0.10% to 0.11%) as reported by two large NRSIs.^106,108 One NRSI¹⁰⁸ reported rates of vertebral artery injury and dural tear of less than 1 percent in for each procedure. One NRSI reported low risk of dysphagia (0% vs. 0.13%)¹⁰⁹ but did not report dysphagia severity. Dysphagia was more common in cervical arthroplasty participants versus ACDF participants (9.4% vs. 6.3%) but severity was not described in one prospective NRSI IDE study using historical ACDF controls.¹²⁰

3.9.3.1.3.6.4. Heterotopic Ossification

Grade 3 or 4 heterotopic ossification (HO), considered clinically relevant HO by most of the trials, may be of concern with cervical arthroplasty. Across five RCTs (N=525 for cervical arthroplasty arm, range 30 to 182), 9.5 percent of participants (range, 1.8% to 12.8%) developed Grade 3 or 4 HO across 24 to 84 months followup.^{61,92,95,97,100} In addition, one FDA IDE NRSI (n=150 in cervical arthroplasty arm) reported rates of grade 3 or 4 HO at 24 months (11.3%; 0.7%, grade 4).¹⁰⁵ Rates of Grade 1 or 2 (or unclear grades) of HO ranged from 0 to 32.7 percent across seven trials (N range for cervical arthroplasty arms, 51 to 201) over 12 to 84 months followup^{60,61,76,79,81,100,101} and was 44 percent at 24 months in the NRSI.¹⁰⁵

3.9.3.1.3.6.5. Device-Related Adverse Events

Device-related adverse event definitions, types of events and adjudication varied across RCTs. Some trials included a range of events such as adjacent-level degenerative joint changes, headache as well as neurological events. Some device-related events may only occur with cervical arthroplasty, others may only occur with ACDF (e.g., nonunion). Some events may not be persistent or serious (e.g., superficial wound infection, dysphagia). Cervical arthroplasty was associated with substantially lower likelihood of device-related events at 24 months (6 RCTs, N=2,167, 4.9% vs. 11%, RR 0.46, 95% CI 0.31 to 0.63, I²=0%).^77,115–119 No difference was seen across two trials at 60 months,^78,102 but results across three trials at >60 months^81,85,97 were inconsistent (Figure 23).

Figure 23

Device-related adverse events: comparison of cervical arthroplasty with ACDF (1-level interventions). ACDF = anterior cervical discectomy and fusion; CI = confidence interval; mos. = months; PL = profile likelihood; SSED = Summary of Safety and Effectiveness (more...)

3.9.3.1.3.6.6. Differential Effectiveness (Heterogeneity of Treatment Effect [HTE])

None of the included trials that compared single-level cervical arthroplasty and ACDF interventions reported differential effectiveness based on patient or other characteristics.

3.9.3.2.1. Fusion

Two RCTs (N=727) (across 4 publications) that compared two-level cervical arthroplasty and ACDF procedures reported fusion success in their ACDF arms.^71,83,94,95 No trials reported short-term fusion success. Two RCTs (N=243) reported intermediate-term fusion success in 92.5 percent (range: 90.5% to 94.0%) of participants.^83,94 Two RCTs (N=196) reported long-term fusion success in 92.6 percent (range: 90.9% to 93.8%) of participants.^71,95 One IDE NRSI¹⁰⁴ comparing a novel cervical arthroplasty versus historical ACDF controls reported pseudarthrosis in 6.5 percent of the ACDF group.

3.9.3.2.2. Pain

3.9.3.2.2.1. Neck Pain

There was moderate-strength evidence of no difference between cervical arthroplasty and ACDF on neck pain (SOE: Moderate).

Two RCTs (N=727)^121,122 that compared cervical arthroplasty with ACDF reported neck pain success (response) defined as postoperative ≥20-point improvement on VAS (0-100 scale). In participants having two-level interventions there were no differences in likelihood of neck pain success between cervical arthroplasty and ACDF in the short term (2 RCTs, N=692, 88% vs. 80.7%, RR 1.10, 95% CI 1.01 to 1.23, I²= 0.8%),^121,122 intermediate term (2 RCTs, N=678, 86.9% vs. 83.3%, RR 1.06, 95% CI 0.98 to 1.15, I²=0%),^121,122 and long term (1 RCT, N=221, 91.2% vs. 81.3%, RR 1.12, 95% CI 1.01 to 1.25)¹²² as estimates were below the threshold for a small effect (Figure 24). There was also no difference long term between cervical arthroplasty and ACDF in the trial using a threshold of ≥10-point improvement for neck pain success that was not included in the meta-analysis (1 RCT, N=269, 86% vs 77.7%, RR 1.11, 95% CI 0.97 to 1.32).⁹⁵

Figure 24

Neck pain success (≥20-point improvement on VAS): comparison of cervical arthroplasty with ACDF (2-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = (more...)

There was no difference in VAS neck pain scores (0-100 scale) between cervical arthroplasty and ACDF short term (3 RCTs, N=764, MD −5.83, 95% CI −12.28 to 0.61, I²=50.3%).^72,94,99 Cervical arthroplasty was associated with a small pain improvement versus ACDF in the intermediate term (4 RCTs, N=707, MD −8.21, 95% CI −13.83 to −4.25, I²=23%)^63,71,94,99 and long term (3 RCTs N=615, MD −8.13, 95% CI −15.18 to −2.97, I²=55.9%)^71,95,99 (Figure 25). One IDE NRSI that compared a novel cervical arthroplasty versus historical ACDF controls reported no differences in mean VAS neck pain intensity at short- or intermediate term (N=352, 1.8 vs. 2.5 at both times, p>0.10).¹⁰⁴

Figure 25

Neck pain scores (0-100): comparison of cervical arthroplasty with ACDF (2-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; FDA = Food (more...)

3.9.3.2.2.2. Arm Pain

There was moderate-strength evidence of no difference between cervical arthroplasty and ACDF on arm pain (SOE: Moderate).

Two RCTs (N=727)^121,122 that compared cervical arthroplasty with ACDF reported arm pain success (response) defined as postoperative ≥20-point improvement on VAS (0-100 scale). Some studies reported arm pain success in both arms. Using conservative estimates (the lower risk ratio), there were no differences in likelihood of arm pain success between cervical arthroplasty and ACDF at short term (2 RCTs, N=692, 70.6% vs. 74.1%, RR 1.0, 95% CI 0.90 to 1.14, I²= 0%),^121,122 intermediate term (2 RCTs, N=678, 71.9% vs.74.0%, RR 1.02, 95% CI 0.92 to 1.14, I²= 0%),^121,122 or long term (1 RCT, N=220, RR 0.94, 95% CI 0.84 to 1.05)¹²² (Figure 26). Estimates and conclusions using the higher risk ratios from the other arm were similar.

Figure 26

Arm pain success (≥20-point improvement on VAS): comparison of cervical arthroplasty with ACDF (2-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = (more...)

Three RCTs (N=792) (in 5 publications)^{63,71,72,94,95} reported arm pain scores (0-100). Some trials reported arm pain scores in both arms. Conservative estimates (using the smaller mean differences) are reported here. There was no difference in VAS arm pain scores (0-100 scale) between cervical arthroplasty and ACDF in the short term (2 RCTs, N=692, MD −3.72, 95% CI −9.53 to 1.62, I²=0%).^72,94 Cervical arthroplasty was associated with a small pain improvement versus ACDF at intermediate term (3 RCTs, N=627, MD −9.95, 95% CI −15.10 to −5.15, I²=0%)^63,71,94 but not long term (2 RCTs N=535, MD −5.08, 95% CI −11.73 to 1.70, I²=1.4%)^71,95 (Figure 27). One IDE NRSI (N=352) that compared a novel cervical arthroplasty versus ACDF using historical controls reported no differences in mean VAS arm pain intensity at short (1.6 vs. 1.7) or intermediate term (1.8 vs. 1.6).¹⁰⁴

Figure 27

Arm pain scores (0-100): comparison of cervical arthroplasty with ACDF (2-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; FDA = Food (more...)

3.9.3.2.3. Function

3.9.3.2.3.1. Neurologic Function

There was moderate-strength evidence of no difference between cervical arthroplasty and ACDF on neurologic function (SOE: Moderate).

Two IDE RCTs (N=727) (in 5 publications)^{71,94,95,121,122} that compared cervical arthroplasty with ACDF reported neurologic success (response), defined as maintenance or improvement (compared with preoperative status) in motor function, sensory function, and deep tendon reflexes. In participants with two-level interventions, there was no difference in likelihood of neurologic success between cervical arthroplasty and ACDF at short term (2 RCTs, N=692, 91.0% vs. 87.9%, RR 1.03, 95% CI 0.96 to 1.10, I²= 0%),^121,122 intermediate term (2 RCTs, N=604, 91.4% vs. 90.6%, RR 0.99, 95% CI 0.93 to 1.07, I²=12.9%)^71,94 or long term (2 RCTs, N=535, 93.2% vs. 84.8%, RR 1.10, 95% CI 1.01 to 1.20, I²=0%; point estimate below the threshold for a small effect)^71,95 (Figure 28). The likelihood of neurological success, based on motor, sensory, and myelopathic gait assessments, was similar for cervical arthroplasty and ACDF in one IDE NRSI (N=352, 100% vs. 97.7%).¹⁰⁴

Figure 28

Neurologic success: comparison of cervical arthroplasty with ACDF (2-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; FDA = Food and Drug (more...)

Mean JOA scores (0-17 scale) were similar following cervical arthroplasty and ACDF at short term (6 months, 15.2 vs. 14.9, p>0.05), intermediate term (15.4 vs. 15.3, p>0.05), and long term (81 months, 15.4 vs. 15.2, p>0.05) in one RCT (N=96).⁹⁹

3.9.3.2.3.2. General Function

There was moderate-strength evidence of no difference between cervical arthroplasty and ACDF on general function (SOE: Moderate).

3.9.3.2.3.2.1. NDI

Two IDE RCTs (N=727) (in 4 publications)^{71,95,121,122} and one IDE NRSI (N=352)¹⁰⁴ that compared cervical arthroplasty with ACDF reported NDI success defined as postoperative NDI score improvement of ≥15 points from baseline. One trial defined NDI success as improvement of ≥30 points from baseline and was not included in the meta-analysis.⁶⁶ Based on the threshold of ≥15 points from baseline, there were no differences between cervical arthroplasty and ACDF (i.e., although statistically significant, the differences between treatments were below the threshold for a small effect) at short term (2 RCTs, N=692, 89.3% vs. 80.0%, RR 1.12, 95% CI 1.04 to 1.22, I²= 0%),^121,122 intermediate term (1 RCT, N=307, 89.2 % vs. 77.9%, RR 1.15, 95% CI 1.03 to 1.27)⁷¹ and long term (2 RCTs, N=535, 84.3% vs. 73.6%, RR 1.16, 95% CI 1.04 to 1.30, I²= 0%)^71,95 (Figure 29). There was no difference in the likelihood of NDI success between cervical arthroplasty and ACDF in one IDE NRSI (N=352, 92.3% vs. 85.5%, p>0.05).¹⁰⁴

Figure 29

NDI success (≥15-point improvement): comparison of cervical arthroplasty with ACDF (2-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; (more...)

One RCT that defined NDI success as improvement of ≥30 points from baseline found a moderately higher likelihood of NDI success following cervical arthroplasty versus ACDF at intermediate term (1 RCT, N=359, 79.3% vs. 53.4%, RR 1.50, 95% CI 1.21 to 1.86).⁶⁶

Four RCTs (N=872) (in 6 publications)^{63,71,72,94,95,99} that compared cervical arthroplasty with ACDF reported NDI scores (0-100, higher score, more limitations). cervical arthroplasty was associated with a small improvement in function based on NDI scores at short (3 RCTs, N=772, MD −5.79, 95% CI −8.44 to −3.21, I²=0%),^72,94,99 intermediate (4 RCTs, N=707, MD −7.69, 95% CI −10.30 to −5.10, I²=0%),^63,71,94,99 and long term (3 RCTS, N=615, MD −7.63, 95% CI −10.64 to −4.52, I²=0%)^71,95,99 (Figure 30).

Figure 30

NDI scores (0-100): comparison of cervical arthroplasty with ACDF (2-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; FDA = Food and Drug (more...)

One IDE NRSI (N=352) that compared a novel cervical arthroplasty versus historical ACDF controls found that cervical arthroplasty was associated with a small improvement in function based on the NDI short term (MD 5.7, means 15.1 vs. 20.8, p<0.05); this was not sustained to intermediate term (MD 2.9, means 14.3 vs. 17.2, p>0.05).¹⁰⁴

3.9.3.2.3.2.2. SF-36 PCS and MCS

Two IDE RCTs (N=727) (in 3 publications)^72,121,122 compared two-level interventions with cervical arthroplasty and ACDF and reported SF-36 PCS and MCS scores (0-100 scale). Success for these component scores was defined as postoperative score improvement of ≥15 points from baseline scores. There was no difference between cervical arthroplasty and ACDF in the likelihood of improved function based on PCS success short term (2 RCTs, N=657, 76.5% vs. 69.3%, RR 1.11, 95% CI 0.88 to 1.46, I²= 72.7%),^121,122 intermediate term (2 RCTs, N=639, 83.7% vs. 79.1%. RR 1.06, 95% CI 0.92 to 1.36, I²=69.7%),^72,121 and long term (1 RCT, N=216, 76.4% vs. 71.0%, RR 1.08, 95% CI 0.91 vs. 1.27)¹²² (Figure 31). Similarly, there were no differences between cervical arthroplasty and ACDF in the likelihood of MCS success at short term (2 RCTs, N=657, 50.3% vs. 45.2%, RR 1.08, 95% CI 0.82 to 1.41, I²= 43.9%),^121,122 intermediate term (2 RCTs, N=639, 62.3% vs. 65.3%, RR 0.98, 95% CI 0.85 to 1.18, I²=0%),^72,121 and long term (1 RCT, N=216, 53.7% vs. 52.7%, RR 1.02, 95% CI 0.79 to 1.31)¹²² (Figure 32).

Figure 31

SF-36 or SF-12 PCS success (≥15-point improvement): comparison of cervical arthroplasty with ACDF (2-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI (more...)

Figure 32

SF-36 or SF-12 MCS success (≥15-point improvement): comparison of cervical arthroplasty with ACDF (2-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI (more...)

Three RCTs (N=792) (in 5 publications)^{63,71,72,94,95} that compared two-level interventions with cervical arthroplasty and ACDF reported SF-36 PCS and MCS scores (0-100 scale). Differences in mean PCS scores did not meet the threshold for a small improvement and were classified as no difference between cervical arthroplasty versus ACDF at short term (2 RCTs, N=692, MD 3.29, 95% CI 0.63 to 6.19, I²=36.6%),^72,94 intermediate term (3 RCTs, N=627, MD 4.80, 95% CI 2.74 to 6.87, I²=0%),^63,71,94 and long term (2 RCTs, N=535, MD 2.32, 95% CI −0.03 to 4.71, I²=0%);^71,95 however, estimates were imprecise (Figure 33). Two RCTs (N=757) reported mean MCS scores which were also not different between groups at short term (1 RCT, N=380, MD 1.00, 95% CI −1.37 to 3.37),⁷² intermediate term (2 RCTs, N=665, MD 1.12, 95% CI −1.07 to 3.29, I²=0%),^66,72 or long term (1 RCT, N=269, MD 2.90, 95% CI −0.25 to 6.05)⁹⁵ (Figure 34). One IDE NRSI (N=352) that compared a novel cervical arthroplasty versus matched historical ACDF controls found no difference in mean SF-36 PCS at short (49.2 vs. 46.4, p<0.05) or intermediate term (49.2 vs. 47.9).¹⁰⁴

Figure 33

SF-36 or SF-12 PCS scores (0-100 scale): comparison of cervical arthroplasty with ACDF (2-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; (more...)

Figure 34

SF-36 or SF-12 MCS scores (0-100 scale): comparison of cervical arthroplasty with ACDF (2-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; (more...)

3.9.3.2.3.2.3. Odom’s Criteria

There was no difference between cervical arthroplasty and ACDF for the likelihood of scoring excellent or good on Odom’s criteria at intermediate term in one RCT (N=62, 96.7% vs. 84.4%, RR 1.15, 95% CI 0.97 to 1.34).⁶³

3.9.3.2.4. Overall Success (Composite)

The FDA IDE trials were required to report on overall success, a composite outcome that included a threshold of ≥15-point NDI improvement from baseline, improvement or maintenance of neurologic status, no serious adverse events and no additional surgical procedures that might be considered “failure” (e.g., removal, revision, supplemental fixation). Cervical arthroplasty was associated with a slightly higher likelihood of overall success short term (2 RCTs, N=693, 73.2% vs. 62.7%, RR 1.19, 95% CI 1.02 to 1.56, I²=56.2%)^94,122 and long term (1 RCT, N=267, 80.4% vs. 62.2%, RR 1.29, 95%CI 1.10 to 1.52).⁷¹ At intermediate term, cervical arthroplasty was also associated with slightly greater likelihood of overall success in two RCTs individually (1 RCT, N=297, 60.1% vs. 31.2%, RR 1.95, 95% CI 1.41 to 2.69 and 1 RCT, N=307, RR 1.21, 95% CI 1.05 to 1.40)^71,94 (Figure 35).

Figure 35

Overall success (composite): comparison of cervical arthroplasty with ACDF (2-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; FDA = Food (more...)

One IDE RCT defined overall success with different NDI success criteria (improvement from baseline of ≥30-points if baseline score was ≥60 or ≥50% if baseline score was <60), required adjudication of adverse events and added radiographic success to the criteria listed for the other IDE trials. Cervical arthroplasty was associated with slightly higher likelihood of overall success long-term versus ACDF (1 RCT, N= 249, 60.8% vs. 34.6%, RR 1.76, 95% CI 1.27 to 2.44).⁹⁵ One IDE NRSI¹⁰⁴ that compared a novel cervical arthroplasty versus historical ACDF controls defined overall success as ≥15-point NDI improvement, maintenance or improvement in neurological status), no serious adverse event (any implant-associated or implant/surgical procedure–associated) and no additional index-level surgical procedure. Authors reported that overall success was more common in cervical arthroplasty participants versus ACDF (N=352, 86.7% vs. 77.1, p<0.05) based on multiple imputation modeling (numerators not reported; effect estimate could not be calculated).

3.9.3.2.5. Quality of Life

None of the included studies reported quality-of-life measures.

3.9.3.2.6. Reoperation

There was low-strength evidence that reoperation is substantially less likely with cervical arthroplasty compared with ACDF at all time points from 24 months and beyond (SOE: Low). Rates of reoperation for ACDF at the index level may be influenced by removal of an existing plate to treat ASD, rather than the indication for reoperation being driven by an issue at the index procedure. This may artificially inflate the reported reoperation rate at the index procedure level for ACDR versus cervical arthroplasty. The clinical relevance of removing the plate as a part of a procedure addressing ASD is minimal.

Reoperation included any additional procedure that involved the index level and was substantially less likely with cervical arthroplasty at all times reported across IDE trials, however estimates were imprecise. Effect estimates were consistent across reported times: up to 24 months (2 RCTs, N=727, 2.8% vs. 9.2%, RR 0.28, 95% CI 0.13 to 0.61, I²=0%),^65,72 36 to 48 months (1 RCT, N=330, 4.0% vs. 15.2%, RR 0.26, 95% CI 0.12 to 0.57),⁶⁶ 60 months (1 RCT, N=330, 4.7% vs. 18.1%, RR 0.26, 95% CI 0.13 to 0.53),⁸⁰ and >60 months (2 RCTs, N=727, 4.4% vs. 15.0%, RR 0.29, 95% CI 0.16 to 0.52, I²=0%)^71,95 (Figure 36). One IDE NRSI that compared a novel cervical arthroplasty versus historical ACDF controls also reported that secondary surgical interventions were less common with cervical arthroplasty (N=352, 2.2% vs. 8.8%).¹⁰⁴

Figure 36

Reoperation at the index level: comparison of cervical arthroplasty with ACDF (2-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; mos. (more...)

Subsequent surgery rates at adjacent levels were similar between cervical arthroplasty and ACDF at 24 months (2 RCTs, N= 727, 1.6% vs. 3.4%, RR 0.51, 95% CI 0.10 to 1.84, I²=19.8%),^72,121 but substantially less common with cervical arthroplasty versus ACDF at 60 months (1 RCT, N=339, 3.4% vs. 11.4%, RR 0.30, 95% CI 0.13 to 0.71)⁸⁰ and >60 months (2 RCTs, N=642, 6.5% vs. 15.1%, RR 0.46, 95% CI 0.25 to 0.80, I²= 0%).^71,95 Across trials, indications for operation at adjacent levels were not consistently described (Figure 37).

Figure 37

Subsequent surgery at adjacent level: comparison of cervical arthroplasty with ACDF (2-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; (more...)

3.9.3.2.7. Harms

Cervical arthroplasty was associated with a slightly lower likelihood of experiencing any adverse event at 24 months based on low-strength evidence (SOE: Low), but there was no difference between procedures at 120 months for WHO Grade 3 or 4 adverse events (SOE: Low). There was insufficient evidence for neurological deficits or events and for mortality (SOE: Insufficient).

All IDE RCTs and one IDE NRSI provided information on adverse events and harms.

3.9.3.2.7.4. Device-Related Adverse Events

Device-related adverse event definitions, types of events and adjudication varied across RCTs. One trial included a range of events such as anatomy/technical difficulty, trauma as well as neurological events while others did not provide specifics. Some device-related events may only occur with cervical arthroplasty, others may only occur with ACDF (e.g., nonunion). Some events may not be persistent or serious (e.g., dysphagia or dysphonia). Two-level cervical arthroplasty was associated with a moderately lower likelihood of device-related events at 24 months compared with ACDF (2 RCTs, N=727, 16.6% vs. 25.6%, RR 0.61, 95% CI 0.38 to 1.01, I²=49.1%)^65,72 but there was no difference between groups at 120 months in one of these trials (N=397, 26.3% vs. 23.4%, RR 1.12, 95% CI 0.80 to 1.59)⁷¹ (Figure 38). When only serious device-related adverse events were considered, as adjudicated by committee or as WHO grade 3 or 4 events, cervical arthroplasty was associated with a substantially lower likelihood of such serious events compared with ACDF at 24 months in one trial (N=397, 1.9% vs. 5.9%, RR 0.33, 95% CI 0.11 to 1.01)⁷² but there was no difference between groups at 120 months in this same trial (RR 0.48, 3.8% vs. 8.1%, 95% CI 0.21 to 1.11)⁷¹ or at 60 months in a second trial (N=330, 4.4% vs. 8.6%, RR 0.52, 95% CI 0.22 to 1.24),⁹⁴ however, the estimates were very imprecise.

Figure 38

Device-related adverse events: comparison of cervical arthroplasty with ACDF (2-level interventions). ACDF = anterior cervical discectomy and fusion; C-ADR = cervical artificial disc replacement (cervical arthroplasty); CI = confidence interval; F/U = (more...)

Device-related adverse events were similar for cervical arthroplasty and ACDF in one IDE NRSI (3.8% vs. 3.5%).¹⁰⁴

3.9.3.2.8. Differential Effectiveness (HTE)

One IDE trial that compared 2-level cervical arthroplasty and ACDF provided subgroup analysis on the presence of radiculopathy alone (N=287) and myelopathy alone or myelopathy with radiculopathy (N=110) for pain, function. and adverse events at 24 and 84 months but did not formally test for interaction.⁷³ Visual inspection of effect estimates and overlap in estimate variability and subgroup estimates suggest no differential effectiveness or harms, although the study may have been underpowered to evaluate this.

3.9.3.3.1. Fusion

One RCT (N=42) reported intermediate-term fusion success in 90.5 percent of participants in the ACDF arm.⁶² This RCT also reported fusion in the cervical arthroplasty arm, but this can be attributed to participant crossover after initial randomization.

3.9.3.3.2. Pain

There was low-strength evidence of no difference between treatment with cervical arthroplasty and ACDF on neck pain (SOE: Low).

There was no difference in median VAS (0 to 10) neck pain scores at 60 months between cervical arthroplasty (3.6, interquartile range [IQR] 3.2 to 4.1) and ACDF (median 3.9, IQR 3.0 to 4.4) at 60 months (p=0.203) in one trial (N=50).⁷⁴ No other pain measures were reported.

3.9.3.3.3. Function

3.9.3.3.4. Quality of Life

None of the included studies reported on quality-of-life measures.

3.9.3.3.5. Harms

There was inadequate evidence to determine the effect of cervical arthroplasty and ACDF on harms or adverse events (SOE: Inadequate).

Two RCTs^62,64 and four NRSIs^{107,110–112} reported harms and adverse events.

3.9.3.3.6. Reoperation and Subsequent Surgery

One RCT (N=53) reported reoperation at the index level in one (4%) cervical arthroplasty and two (7.1%) ACDF participants between 12 and 36 months (RR 0.56, 95% CI 0.05 to 5.81).⁶⁴ A second trial (N=83) reported that no participants in either group required reoperation at the index level through 36 months.⁶² One NRSI did not provide adjusted effect estimates but reported the proportions of cervical arthroplasty and ACDF patients who required reoperation at the index level at 12 months (1.7% vs. 2.4%) and 24 months (0% vs. 3.6%) and subsequent surgery at adjacent levels at 12 months (1.7% vs. 2.4%) and 24 months (3.3% vs. 5.1%).¹¹⁰ Across the two NRSIs that did attempt to control for confounding (propensity score adjusted analyses), over the first 30 postoperative days, 0.4 percent of cervical arthroplasty versus 1.0 percent of ACDF underwent any reoperation (not further specified) in one study (N=1,368)¹¹² and in the second study, 2.8 versus 1.0 percent had a revision surgery, 0.4 versus 0.2 percent had a drainage/evacuation, and no patient had a hardware removal in the other study (N=1,014).¹¹¹ At 12 months in the latter study, the proportion of patients requiring revision surgery rose to 10.7 versus 7.1 percent; the need for drainage/evacuation (0.8% for both) and hardware removal (0.2% for both) remained low.

3.9.3.3.7. Differential Effectiveness (HTE)

None of the included trials that compared 1-, 2-, or 3 level cervical arthroplasty and ACDF interventions reported differential effectiveness based on patient or other characteristics.

3.10.1.3.1. Fusion

There was moderate-strength evidence of no difference in fusion rates between standalone cages versus plate and cage in participants undergoing ACDF (SOE: Moderate).

Almost all participants who underwent ACDF with either a standalone cage or with a traditional plate and cage (N=515) experienced fusion at 12 months (4 RCTs, N=178, 94% vs. 97%, RR 0.99, 95% CI 0.92 to 1.06, I²=0%), 24 months (2 RCTs, N=150, 95% vs. 95%, RR 1.00, 95% CI 0.93 to 1.08, I²=0%) and 36 months (2 RCTs, N=187, 100% vs. 100%, RR 1.00, 95% CI 0.97 to 1.03, I²=0%) (Figure 39). This was true when fusion was limited to one level or involved multilevel fusion. One trial did not report fusion as an outcome.¹³¹ (SOE: Moderate)

Figure 39

Fusion, standalone cage versus traditional plate and cage. CSLP = cervical spine locking plate; CI = confidence interval; PEEK = polyetheretherketone; PL = profile likelihood; ROI-C = ROI-C implant system; Zero-P = zero-profile

3.10.1.3.2. Pain

There was low-strength evidence of no difference between standalone cages versus plate and cage on arm pain (SOE: Low), with inadequate evidence to determine the benefits and harms of the two approaches on neck pain (SOE: Insufficient).

Four RCTs (N=230) reported changes in overall pain (pain location not specified) or neck pain using a visual analogue scale (VAS: 0-10 or 0-100) across various followup times ranging from less than 3 months to 24 months (Figure 40). Although neck pain was moderately, though not statistically greater at less than 3 months (MD −0.90, 95% CI −1.29 to 0.73) with a plate and cage compared with a standalone cage, the opposite was true at 6 months , MD 0.64, 95% CI −0.66 to 2.17). When pooled analysis was limited to trials of single-level disease, there were no differences in neck pain between standalone cage and plate and cage (Appendix F, Figure F-6).

Four RCTs (N=186) reported changes in arm pain using a visual analogue scale (VAS: 0-10 or 0-100) across various followup times. There were no differences in arm pain after ACDF between use of a standalone cage and a plate and cage at any time point from less than 3 months (MD −0.24, 95% CI −1.55 to 1.12) to 24 months (MD 0.20, 95% CI −0.09 to 0.49) (Figure 41). When analyses were limited to trials of single-level disease, there remained no difference in arm pain between fusion methods (Appendix F, Figure F-7).

Figure 40

Neck/unspecified pain after ACDF. ACDF = anterior cervical discectomy and fusion; CI = confidence interval; PEEK = polyetheretherketone; PL = profile likelihood; ROI-C = ROI-C implant system; SD = standard deviation; Zero-P = zero-profile

Figure 41

Arm pain following ACDF. ACDF = anterior cervical discectomy and fusion; CI = confidence interval; CSLP = cervical spine locking plate; PEEK = polyetheretherketone; PL = profile likelihood; ROI-C = ROI-C implant system; SD = standard deviation; Zero-P (more...)

3.10.1.3.3. Function

3.10.1.3.3.1. Neurologic Function

There was low-strength evidence of no difference between standalone cages versus plate and cage in neurologic function (SOE: Low).

Five RCTs (N=424) reported changes on the JOA (lower score = worse disability, score 0 to 17) after ACDF using a standalone cage or a plate and cage (Figure 42). At less than 3 months, pooled analysis of two trials indicated moderately greater, although not statistically significant, JOA scores with a standalone cage versus a plate and cage (MD 2.63, 95% CI −3.86 to 9.29), this effect is driven by 1 of 2 trials, while the other trial found no effect. At longer followup times, there were no differences between treatments on JOA scores.

Figure 42

JOA scores following ACDF. ACDF = anterior cervical discectomy and fusion; CI = confidence interval; CSLP = cervical spine locking plate; JOA = Japanese Orthopaedic Association Scale; PEEK = polyetheretherketone; PL = profile likelihood; ROI-C = ROI-C (more...)

3.10.1.3.3.2. General Function

There was low-strength evidence of no difference between standalone cages versus plate and cage in general function (SOE: Low).

Six RCTs (N=472) reported changes on the NDI (higher score = worse disability, 0-50 raw score or 0% to 100%) following ACDF with either a standalone cage or a plate and cage (Figure 43). With the exception of less than 3 months timepoint, there were no differences between ACDF with a standalone cage or plate and cage on NDI scores at other timepoints. At less than 3 months, study findings varied and although the pooled estimate slightly favors the standalone cage (MD −5.39, 95% CI −9.91 to 5.19), it is driven by the largest of the three studies and should interpreted with caution.

Figure 43

NDI scores following ACDF. ACDF = anterior cervical discectomy and fusion; CI = confidence interval; NDI = Neck Disability Index; PEEK = polyetheretherketone; PL = profile likelihood; ROI-C = ROI-C implant system; SD = standard deviation; Zero-P = zero-profile (more...)

Additionally, one trial (N=41) reported no difference at 24 months between a standalone zero-profile device (Zero-P) and a plate and cage on the German version of the Neck Pain Disability Index (25.8% vs. 22.2%, p-value not reported).¹²⁵

One RCT (N=46) reported no difference between a standalone cage and plate and cage at 24 months on the Odom’s criteria (Excellent: 46% vs. 55%; Good: 54% vs. 45%; Fair: 0% vs. 0%; Bad: 0% vs. 0%),¹³⁰ while another trial (N=41) reported the mean Odom’s Criteria at 24 months was 3.2 with a standalone cage compared with 3.5 with plate and cage (p-value not reported).¹²⁵ A third trial (N=115) reported there were no differences between standalone cage versus plate and cage in ratings of “excellent” and “good” overall patient satisfaction (Excellent: 44% vs. 47%, p=0.763; Good: 33% vs. 29%, p=0.835; Fair: 23% vs. 24%, p=0.692; Poor: 0% vs. 0%, p=1.0) at 36 months.¹²⁴

3.10.1.3.4. Quality of Life

There was low-strength evidence of no difference between standalone cages versus plate and cage in quality of life (SOE: Low).

One RCT (N=40) reported no differences in quality of life as assessed with the Veteran’s RAND 12-Item Health Survey between treatment with a standalone cage versus a plate and cage at 6 weeks and at 12 months, although participants treated with a standalone cage reported better scores at 6 months postoperatively (38.38 vs. 26.27, p=0.033).¹³²

Five RCTs (N=253) assessed swallowing before and after treatment with a standalone cage versus a plate and cage with mixed results.^{125–128,132} Two trials used the Swallowing Quality of Life questionnaire,^127,132 two trials rated severity of dysphagia symptoms as “None”, “Mild”, “Moderate”, and “Severe”^125,128 and one trial used the Eating Assessment Tool.¹²⁶ No trial reported differences in dysphagia scores between treatments beyond 3 months postoperatively. One trial reported worse dysphagia scores with plate and cage immediately postoperatively, at 1 month, and at 3 months but no difference at 12 months.¹²⁸ Another trial reported worse scores with plate and cage at 6 weeks but no differences at 6 and 12 months.¹³² There were no differences between dysphagia scores at any time from the postoperative period to 12 month in one RCT¹²⁶ and no differences at 36 months (only time reported) in another trial.¹²⁷ One trial reported no patient rated dysphagia as “moderate” or “severe” with either treatment¹²⁵ and no study reported that dysphagia required medical intervention (e.g., return to the operating room, percutaneous endoscopic gastrostomy tube placement).

One RCT (N=54) rated high risk of bias found no differences on the Voice Handicap Index between treatment with a standalone cage versus plate and cage from discharge to 12 months.¹²⁶

3.10.1.3.5. Harms

There was low-strength evidence of no difference between standalone cage versus plate and cage on adjacent-level ossification (SOE: Low), while evidence for subsidence and other adverse events was inadequate (SOE: Insufficient).

Seven RCTs (N=518) reported adverse events.^{124,127–132} Three trials reported substantially less adjacent-level ossification development with a standalone cage than with plate and cage (N=239, 8% vs. 27%, RR 0.25, 95% CI 0.12 to 0.52, I²=8%). The change in adjacent-level ossification development severity grade (0=no ossification, 3=severe ossification) was reported in one study and favored treatment with the standalone cage (0.208 vs. 0.818, p=0.001).¹³⁰ (SOE: Low) However, no patient required reoperation at 36 months in two trials;^124,127 reoperation rates were not reported in the third trial.¹³⁰

One RCT (N=46) reported a small, but not statistically significant difference in subsidence (loss of disc height) rates with a standalone cage compared with a plate and cage at 12 months (12.5% vs. 9.1%, RR 1.38, 95% CI 0.25 to 7.48) and at 24 months (16.7% vs. 13.6%, RR 1.22, 95% CI 0.31 to 4.87).¹³⁰

One trial (N=104) reported few total complications (N=11) in 24 months that included one nerve injury (2%) and no cerebrospinal fluid leaks (0%) with the standalone cage compared with two nerve injuries (4%) and one cerebrospinal fluid leak (2%) with the plate and cage (p=0.999; p=1.00, respectively).¹²⁹ One trial (N=90) reported one (2%) incidence of loosening of the internally fixed implant with the standalone cage versus three (7%) with plate and cage (p=0.333).¹³¹ Another trial (N=40) reported participant treated with a standalone cage experienced a screw loosening, interbody subsidence, and C-5 fracture with revision surgery under consideration at trial publication.¹³² The same trial also reported one participant treated with a plate and cage experienced screw fracture, pseudarthrosis and underwent posterior fusion and decompression 14 months after the primary surgery.

3.10.2.3.1. Fusion

There was low-strength evidence of a greater likelihood of fusion with a PEEK cage compared with a titanium or titanium-coated PEEK cage (SOE: Low)

Three RCTs (N=217) reported ACDF fusion rates at different followup times that were not different between titanium and PEEK cages or that favored PEEK cages.

One trial reported that at a mean of 99.7 months (range 86 to 116 months) all participants (N=60) achieved fusion of their 3-level disease with both the titanium cage and with the PEEK cage (87/87 levels vs. 93/93 levels).¹³³ However, followup was not available for 25 percent of the original participants. A second trial (N=53) reported a lower likelihood of fusion with the titanium cage (32/37 levels, 86.5%) versus the PEEK cages (34/34 levels, 100%, p=0.0335) after 24 months.¹³⁴ The third RCT (N=104) reported a large difference in the likelihood of complete fusion that favored the PEEK cage with complete fusion achieved in 26 of 59 titanium-covered PEEK cages implanted (44.1%) compared with 75 of 85 PEEK cages implanted (88.2%) at 12 months (p<0.001).¹³⁵ Partial fusion was achieved by 55.9 percent of participants with titanium-covered PEEK cages and 11.76 percent of participants with PEEK cages.¹³⁵ There were no instances of an absence of fusion.¹³⁵

3.10.2.3.2. Function

3.10.2.3.3. Quality of Life

No studies reported quality of life outcomes.

3.10.2.3.4. Harms

Evidence was inadequate to determine the effect of a PEEK cage versus a titanium cage on subsidence or other adverse events (SOE: Insufficient).

One RCT (N=104) found no difference between a titanium-coated PEEK implant and a PEEK implant on the incidence of subsidence in 166 levels (20.6% vs. 21.4%, p=0.875).¹³⁵ However, subsidence was reported with 34.5% of titanium cages (87 levels) compared with 5.4% of PEEK cages (93 levels) in a second RCT (N=60, p<0.05)¹³³ and 16.2% of 37 levels versus 0% of 34 levels in a third RCT (N=53, p<0.001).¹³⁴ All three trials defined subsidence similarly (≥ 3 mm of interspace collapse). It is unclear the reason for the difference in study findings; possibilities include the cage materials (a titanium-coated PEEK cage may perform differently than a titanium cage) and the duration since ACDF (12 months in the trial that found no difference versus 24 months and 99.7 months in the other two trials) (SOE: Insufficient).

One RCT (N=53) reported that after 24 months, there were no neurovascular injuries and no revision surgeries with either the titanium cage or the PEEK cage, but that one patient, who received the titanium cage, experienced a hematoma that was removed the day after surgery.¹³⁴ One RCT (N=60) reported that at a mean of 99.7 months two patients treated with a titanium cage experienced cage dislocation but were asymptomatic.¹³³

3.10.3.3.1. Fusion

There was inadequate evidence to determine the comparative benefits and harms of autograft, allograft, or other osteogenic material versus any other osteogenic material on fusion (SOE: Insufficient).

Six RCTs (N=534) assessed ACDF with autograft, allograft, or other materials (e.g., hydroxyapatite, calcium sulphate) and found no differences between materials in achievement of spinal fusion (Table 3). Fusion rates for all materials were high for all trials but only one randomized study was available for each comparison.

Table 3

Fusion with ACDF using various osteogenic materials.

3.10.3.3.2. Pain

There was inadequate evidence to determine the comparative benefits and harms of autograft, allograft, or other osteogenic material versus any other osteogenic material on neck or arm pain (SOE: Insufficient).

Five RCTs (N=440) assessed neck and arm pain using a VAS or a numerical (pain) rating scale (Tables 4 and 5). One small trial (N=27) reported a moderately greater decrease in neck pain 12 months after ACDF with a local graft and titanium cage than with allograft and titanium cage (MD −6.15 vs. −5.09, p<0.05).¹³⁶ Another trial (N=20) found a moderate, though not statistically significant, improvement in neck pain with BMP-2 and allograft ring versus iliac crest bone graft and an allograft ring on a 20-point numerical rating scale (MD 13.0 vs. MD 9.0, p>0.05).¹⁴⁰

One trial (N=27) also found a substantially greater decrease in arm pain with local graft and a titanium cage compared with allograft and the same cage (MD −7.24 vs. MD −4.55, p<0.05)¹³⁶ (Table 5). However, these results should be interpreted with caution due to the trial’s small sample size. One RCT (N=26) reported a substantially greater reduction in arm pain at 24 months with BMP-2 and allograft ring compared with iliac crest bone graft and allograft ring on a 20-point numerical rating scale (−14 vs. −8.5, p<0.03).¹⁴⁰ However, as above, these results should be interpreted with caution due to the small sample size. One RCT (N=244) found that ACDF with i-Factor (bone graft made of a peptide bound to an inorganic bone mineral) and an allograft ring was associated with improved VAS arm pain scores at 24 months (1.56 v s. 1.95, p=0.0306) compared with local graft and an allograft ring.¹³⁷ However, this small difference in scores is below the threshold for a small effect and may not be clinically meaningful. One RCT (N=77) found a small, although not statistically significant, improvement in arm pain at 12 months with hydroxyapatite, demineralized bone matrix and a PEEK cage compared with β-tricalcium phosphate, hydroxyapatite and a PEEK cage (VAS: MD −4.2 vs. MD −3.6, p=0.27).¹⁴¹

There were no differences in neck or arm pain with other comparisons.

Table 4

Neck pain with ACDF using various osteogenic materials.

Table 5

Arm pain with ACDF using various osteogenic materials.

3.10.3.3.3. Function

3.10.3.3.3.1. Neurologic Function

There was inadequate evidence to determine the comparative benefits and harms of autograft, allograft, or other osteogenic material versus any other osteogenic material on neurologic function (SOE: Insufficient).

Four RCTs (N=436) reported changes in neurological status after ACDF (Table 6). One trial (N=100) found no differences between use of biphasic calcium phosphate ceramic plus a PEEK cage compared with iliac crest bone graft plus a peek cage on JOA score, or JOA recovery rate at 6 months post ACDF.¹³⁹ One trial (N=66) reported no difference between calcium sulphate plus demineralized bone matrix plus a PEEK cage versus autogenous iliac cancellous bone plus a PEEK cage in JOA scores at 24 months.¹³⁸ One trial (N=26) reported neurologic success (i.e., maintenance or improvement in sensory and motor function) in all remaining participants at 24 months,¹⁴⁰ while another trial (N=244) reported that almost all participants (94.87% vs. 93.70%) experienced neurologic success, also at 24 months.¹³⁷

Table 6

Neurologic function with ACDF using various osteogenic materials.

3.10.3.3.3.2. General Function

There was inadequate evidence to determine the comparative benefits and harms of autograft, allograft, or other osteogenic material versus any other osteogenic material on general function (SOE: Insufficient).

Four RCTs (N=374) assessed post ACDF neck disability with the NDI (Table 7). One RCT (N=244) found that treatment with i-Factor plus an allograft ring in ACDF resulted in slightly, though not statistically significant, improvement on NDI endpoint scores at 24 months compared with local graft and allograft ring (22.33 vs. 25.66, p=0.5607).¹³⁷ One small trial (N=26) reported moderately greater improvement on the NDI after 24 months with BMP-2 and allograft ring compared with iliac crest bone graft and allograft ring (52.7 vs. 36.9, p<0.03).¹⁴⁰ Another small trial (N=27) reported moderately greater improvement on NDI scores after 12 months with local graft plus a titanium cage versus allograft plus titanium cage (MD 56.5 vs. MD 41.4, p<0.05).¹³⁶ There was no difference in improvement in NDI scores with hydroxyapatite/demineralize bone matrix plus PEEK cage versus β-tricalcium phosphate/hydroxyapatite plus PEEK cage at 12 months.¹⁴¹

Three RCTs (N=357) assessed general function using the SF-36 or the 2-item SF-12 (Table 7). Two trials found no difference in function on the SF-36 after ACDF using an allograft ring with either i-Factor or local graft¹³⁷ or using an allograft with either BMP-2 or an iliac crest bone graft.¹⁴⁰ One small trial (N=27) reported moderately better function at 12 months using the 2-item SF-12 with local graft plus a titanium cage compared with the same cage and allograft infused with the participant’s blood (MD 48.7 vs. 65.9, p<0.05).¹³⁶ However, care should be used in interpreting these results due to the small study sample size.

Table 7

General function with ACDF using various osteogenic materials.

3.10.3.3.4. Harms

There was low-strength evidence that the use of BMP-2 in cervical spine fusion is associated with increased complications compared to the use of no BMP-2 (SOE: Low), while evidence was inadequate to determine the comparative harms of other osteogenic materials (SOE: Insufficient).

Four RCTs (N=520) and 2 NRSI studies (N=944) reported harms with ACDF using various graft materials (Table 8). There were few differences between treatments reported in the RCTs in the likelihood of various harms. One trial (N=319) reported a moderately greater likelihood of experiencing a new radiculopathy with an allograft ring with local graft than with i-Factor (13.66% vs. 25.00%, p=0.0142) but there were no differences in new intractable neck pain or progression of neuropathy.¹³⁷ One trial (N=100) reported a shorter hospital stay with a biphasic calcium phosphate ceramic combined with a PEEK cage compared with a PEEK cage with iliac crest bone graft.¹³⁹ Reasons for the difference in hospital stay were not provided.

Two retrospective NRSI of BMP-2 compared with no BMP-2 in ACDF (N=944) reported a greater likelihood of heterotopic ossification (78.6% vs. 59.2%, p<0.001)¹⁴² and complications associated with neck swelling¹⁴³ with the use of BMP-2 (Table 8). In one NRSI, participants were 10 times more likely to have a neck swelling complication if BMP-2 was used in anterior cervical fusion, even after controlling for potential confounding variables (e.g., age, gender, presence of myelopathy, levels fused, smoking).¹⁴³

Table 8

Adverse events with ACDF using various graft materials.

3.12.3.3.1. Systematic Review Evidence

A 2013 systematic review that assessed the prognostic utility of preoperative MRI for neurologic recovery after surgery included 22 studies (N=1,508).¹⁴⁴ The included studies evaluated preoperative MRI in patients undergoing cervical disc surgery using a posterior approach (k=7), ACDF (k=5), mixed approaches (k=9), or an unspecified procedures (k=1) over followup ranging from 1.5 to 60.6 months (mean 27.8; standard deviation 4.6 months). The majority of patients in the included studies were male (mean proportion of females: 27.1%), and the mean age (from 20 studies reporting age) was 57.4 (standard deviation, 1.0) years. Heterogeneity of study designs, methods, and outcomes (JOA in 17 studies, Nurick grade in 5 studies, NDI in one study, and Neurosurgical Cervical Spine Score in one study) of the included studies precluded pooling of study findings, and the mixed results were reported narratively. Presence of multisegmental T2-weighted increased signal intensity (ISI) was associated with worse functional outcomes in five studies, not associated with outcomes in four other studies, and lack of T2-weighted ISI was associated with better outcomes in three studies; qualitative classification of T2-weighted ISI was associated with poorer functional status in six studies, not associated with functional outcomes in one study, and lack of T2-weighted ISI associated with better outcomes in one study. Snake-eye appearance on axial T2-weighted MRI, ISI in gray and white matter, and increased SIR were associated with poorer surgical outcomes in one study each.

3.12.3.3.2. Primary Study Evidence

We identified four relevant studies (N=326) that were not included in the systematic review,^{156,157,159,160} as well as 13 studies (in 15 publications) that were published subsequent to the review search dates.^{145–155,158,161–163} Of these studies, two assessed presence of segmental abnormalities (endplate abnormalities, modic changes, and Cobb angle/loss of lordosis),^146,147,152 six assessed qualitative differences in ISI intensity,^{145,149–151,154,157} three assessed SIR,^148,153,155 one evaluated presence or absence of signal changes,¹⁵⁹ two evaluated diffusion tensor tractography grading,^158,162 one (in 2 publications) evaluated diffusion-based spectrum imaging,^161,162 one evaluated a radiomic-based extra tree model,¹⁶³ one evaluated the size of the transverse area at the compression site,¹⁶⁰ and one evaluated size, extent, and qualitative intensity.¹⁵¹ The study (N=55) that assessed the size of the transverse area reported significant associations with postoperative JOA scores (r=0.298) and with JOA recovery (r=0.295) (both p<0.05).¹⁶⁰ The study (N=56) that evaluated size, extent, and intensity of ISI reported no association of size or extent of ISI with functional outcomes;¹⁵¹ one other study of qualitative imaging signal intensity also reported no association of intensity changes with recovery (mJOA score ≥16, RR 1.71; 95% CI 0.90 to 3.24),¹⁴⁵ while four studies (N=714) did find qualitative intensity associated with reduced recovery ratio, lower likelihood of optimal surgical outcome, or change in JOA or NDI scores.^{149–151,154} One study (N=52) reported improved JOA recovery rate (54.3% vs. 27.3%) in patients without ISI compared to those with ISI.¹⁵⁶ Another study (N=146) that assessed presence or absence of imaging signal changes reported that patients without imaging signal changes were more likely to have improvement in Nurick grade (OR 5.1; 95% CI, 1.87 to 25.1); however, there was no difference between patients without imaging signal changes and those with only T2-weighted signal changes.¹⁵⁹ Another study (N=73) found that the combination of T1-weighted hypointensity and T2-weighted hyperintensity was associated with poorer JOA recovery than T2-weighted hyperintensity alone or no ISI changes (JOA recovery 48% vs. 19% vs. 60.7%; T1- and T2-weighted ISI changes vs. T2-weighted ISI change only, p=0.0259).¹⁵⁷ Two studies of SIR (N=220) reported increased T2-weighted SIR associated with JOA recovery;^148,155 one study (N=148)¹⁵³ reported no association between T2-weighted SIR and outcomes, while lower T1-weighted SIR was associated with poorer neurological outcomes assessed with the JOA. One study (N=129)¹⁵⁸ found that diffusion tensor tractography grading using MRI images was associated with JOA score changes (r= −0.813, p<0.001) and JOA recovery (r= −0.429, p<0.001), while conventional MRI ISI grading was associated with JOA score changes (r= −0.674, p<0.001) but not with JOA recovery (r= −0.197, p=0.058). However, another study (N=42) comparing diffusion-based spectrum imaging to diffusion tensor grading reported that no diffusion tensor metrics were associated with neurological (mJOA) or general function (SF-36, NDI, and Myelopathy Disability Index) outcomes.¹⁶² The study found that preoperative diffusion-based spectrum imaging intra-axonal axial diffusivity and anisotropic fraction correlated with improved mJOA scores (r=0.37, p=0.02 and r=0.34, p=0.03, respectively).¹⁶² Another analysis of most of these same patients (N=50)¹⁶¹ compared diffusion-based spectrum imaging to clinical features and found greater prognostic utility with diffusion-based spectrum imaging (area under the curve [AUC] 75.3%) and the combination of diffusion-based spectrum imaging with clinical features (AUC 98.0%) than with assessment of clinical features alone (AUC 59.4%) for mJOA scores. The study reported similar findings for the prognostic utility of diffusion-based spectrum imaging (AUC 54.6%) or the combination of diffusion-based spectrum imaging and clinical features (AUC 65.3%) versus clinical features alone (AUC 48.8%) for NDI.

One study (N=151) evaluated a novel radiomic-based extra tree model of MRI data for predicting neurological outcomes following surgery for CSM.¹⁶³ The study reported that their radiomic-based model (AUC 75%) and the combination of their radiomic-based model with clinical assessment (AUC 71%) were superior to radiological assessment (AUC 43%) and the combination of radiological and clinical assessment (AUC 40%) for predicting neurologic recovery assessed using mJOA.

One study (N=121) reported a novel classification system for reporting loss of cervical lordosis following laminoplasty was predicted by an interplay of preoperative Cobb angle, T1 slope, and dynamic extension reserve.¹⁵² One study (N=861) reported Modic changes, defined as “subchondral vertebral bone marrow lesions of the endplate” on preoperative MRI and found that while modic changes were associated with greater postoperative disability, modic changes were also associated with older age, greater number of levels fused, and a longer duration of symptoms.¹⁴⁶

Comparing findings across studies was difficult due to the various study methods used (e.g., different type and basis of classification of T2 weighted ISI [single segment, multisegment, L2 classification, Q3 classification, SIR], different outcomes assessed [JOA, NDI, Nurick grade], and different methods to analyze the data [correlation, linear regression, multivariable regression, Student’s t test]). Preoperative MRI also preceded different types of surgery (e.g., ACDF, laminoplasty, posterior-anterior decompression), which reduces the generalizability of findings.

3.12.3.3.3. Synthesis of Systematic Review and Primary Study Findings

There was low-strength evidence that multisegmental T2-weighted-increase signal intensity and sharp T2-weighted-increased signal intensity on preoperative MRI was associated with poorer neurologic outcomes (SOE: Low); there was also low-strength evidence that increased SIR of preoperative MRI was associated with poorer neurologic recovery (SOE: Low)

In total, presence of ISI was associated with poorer neurologic outcomes (e.g., JOA recovery, Nurick grade, NDI) in 7 studies and absence of ISI was associated with better neurologic outcomes (e.g., JOA, Nurick grade) in 4 studies but was not associated with changes in neurologic outcomes in 5 studies. Qualitative grading (increased intensity) of ISI was associated with worse neurologic outcomes (e.g., JOA, NDI) in 11 studies, absence of T2-weighted intensity associated with a better neurologic outcome (Nurick grade) in 1 study, and not associated with neurologic outcomes in 3 studies. Higher SIR was associated with poorer recovery in 3 studies (AUCs ranged from 78.6% to 87.3% in the two studies that reported accuracy results); one study reported lower SIR on T1 weighted associated with poorer neurological outcomes (e.g., JOA), while T2-weighted SIR was not associated with outcomes. One study reported that diffusion tensor tractography grading was more closely associated with neurological outcomes and recovery (e.g., JOA) than conventional ISI grading; however, another study found no association of diffusion tensor grading with neurological outcomes. One study of diffusion-based spectrum imaging found the imaging modality superior to diffusion tensor grading and assessment using clinical features. One study found a novel radiomic-based extra tree model to be superior to both radiological and clinical assessment (SOE for ISI and SIR: Low).