Predictors of long-term clinical outcomes in adult patients after lumbar total disc replacement: development and validation of a prediction model

Domagoj Coric MD1, Jack Zigler MD2, Peter Derman MD, MBA2, Ernest Braxton MD, MBA3, Aaron Situ MSc, MQF4, and Leena Patel PhD4
View More View Less
  • 1 Atrium Musculoskeletal Institute, Spine Division, Carolina Neurosurgery and Spine Associates, Charlotte, North Carolina;
  • | 2 Department of Spinal Surgery, Texas Back Institute, Plano, Texas;
  • | 3 Department of Neurological Surgery, Vail Health Vail-Summit Orthopaedics and Neurosurgery, Vail, Colorado; and
  • | 4 Value & Evidence, EVERSANA Life Science Services LLC, Burlington, Ontario, Canada
Full access

OBJECTIVE

Long-term outcomes of single-level lumbar arthroplasty are understood to be very good, with the most recent Investigational Device Exemption (IDE) trial showing a < 5% reoperation rate at the close of the 7-year study. This post hoc analysis was conducted to determine whether specific patients from the activL IDE data set had better outcomes than the mean good outcome of the IDE trial, as well as to identify contributing factors that could be optimized in real-world use.

METHODS

Univariable and multivariable logistic regression models were developed using the randomized patient set (n = 283) from the activL trial and used to identify predictive factors and to derive risk equations. The models were internally validated using the randomized patient set and externally validated using the nonrandomized patient set (n = 52) from the activL trial. Predictive power was assessed using area under the receiver operating characteristic curve analysis.

RESULTS

Two factors were significantly associated with achievement of better than the mean outcomes at 7 years. Randomization to receive the activL device was positively associated with better than the mean visual analog scale (VAS)–back pain and Oswestry Disability Index (ODI) scores, whereas preoperative narcotics use was negatively associated with better than the mean ODI score. Preoperative narcotics use was also negatively associated with return to unrestricted full-time work. Other preoperative factors associated with positive outcomes included unrestricted full-time work, working manual labor after index back injury, and decreasing disc height. Older age, greater VAS–leg pain score, greater ODI score, female sex, and working manual labor before back injury were identified as preoperative factors associated with negative outcomes. Preoperative BMI, VAS–back pain score, back pain duration ≥ 1 year, SF-36 physical component summary score, and recreational activity had no effect on outcomes.

CONCLUSIONS

Lumbar total disc replacement for symptomatic single-level lumbar degenerative disc disease is a well-established option for improving long-term patient outcomes. Discontinuing narcotics use may further improve patient outcomes, as this analysis identified associations between no preoperative narcotics use and better ODI score relative to the mean score of the activL trial at 7 years and increased likelihood of return to work within 7 years. Other preoperative factors that may further improve outcomes included unrestricted full-time work, working manual labor despite back injury, sedentary work status before back injury, and randomization to receive the activL device. Tailoring patient care before total disc replacement may further improve patient outcomes.

ABBREVIATIONS

AUC = area under the receiver operating characteristic curve; DDD = degenerative disc disease; FDA = Food and Drug Administration; IDE = Investigational Device Exemption; ODI = Oswestry Disability Index; PCS = physical component summary; ROM = range of motion; SAE = serious adverse event; TDR = total disc replacement; VAS = visual analog scale.

OBJECTIVE

Long-term outcomes of single-level lumbar arthroplasty are understood to be very good, with the most recent Investigational Device Exemption (IDE) trial showing a < 5% reoperation rate at the close of the 7-year study. This post hoc analysis was conducted to determine whether specific patients from the activL IDE data set had better outcomes than the mean good outcome of the IDE trial, as well as to identify contributing factors that could be optimized in real-world use.

METHODS

Univariable and multivariable logistic regression models were developed using the randomized patient set (n = 283) from the activL trial and used to identify predictive factors and to derive risk equations. The models were internally validated using the randomized patient set and externally validated using the nonrandomized patient set (n = 52) from the activL trial. Predictive power was assessed using area under the receiver operating characteristic curve analysis.

RESULTS

Two factors were significantly associated with achievement of better than the mean outcomes at 7 years. Randomization to receive the activL device was positively associated with better than the mean visual analog scale (VAS)–back pain and Oswestry Disability Index (ODI) scores, whereas preoperative narcotics use was negatively associated with better than the mean ODI score. Preoperative narcotics use was also negatively associated with return to unrestricted full-time work. Other preoperative factors associated with positive outcomes included unrestricted full-time work, working manual labor after index back injury, and decreasing disc height. Older age, greater VAS–leg pain score, greater ODI score, female sex, and working manual labor before back injury were identified as preoperative factors associated with negative outcomes. Preoperative BMI, VAS–back pain score, back pain duration ≥ 1 year, SF-36 physical component summary score, and recreational activity had no effect on outcomes.

CONCLUSIONS

Lumbar total disc replacement for symptomatic single-level lumbar degenerative disc disease is a well-established option for improving long-term patient outcomes. Discontinuing narcotics use may further improve patient outcomes, as this analysis identified associations between no preoperative narcotics use and better ODI score relative to the mean score of the activL trial at 7 years and increased likelihood of return to work within 7 years. Other preoperative factors that may further improve outcomes included unrestricted full-time work, working manual labor despite back injury, sedentary work status before back injury, and randomization to receive the activL device. Tailoring patient care before total disc replacement may further improve patient outcomes.

In Brief

Surgeons and researchers used statistical modeling to identify characteristics of patients with lumbar degenerative disc disease that may predict better outcomes after total disc replacement than the average 7-year outcomes seen in a clinical trial. Preoperative narcotics use was associated with less functional improvement than the 7-year average and reduced chance of return to unrestricted full-time work. Tailoring patient care before lumbar total disc replacement may further improve patient outcomes.

Disabling chronic low-back pain due to lumbar degenerative disc disease (DDD) is prevalent and constitutes a significant burden to society in terms of high narcotics use, healthcare resource use, and missed worked days.1 The estimated economic impact of low-back pain in the United States exceeds $100 billion per year.2 Total disc replacement (TDR) effectively treats symptomatic lumbar DDD in appropriately selected patients for whom at least 6 months of conservative care has failed.3 Clinical studies show that TDR is associated with improved clinical outcomes compared with those of lumbar fusion and nonsurgical treatments.4–10

Multiple factors have been identified as positive prognostic indicators associated with TDR. Less time off work before surgery was predictive of good outcomes after TDR in one study.11 Other studies have identified positive preoperative, intraoperative, and postoperative predictive factors such as high-grade disc degeneration, Modic changes, low disc height, increased postoperative disc height, short preoperative duration of low-back pain, and fewer fear-avoidance beliefs about work.10,12,13 In contrast, comorbidities, low education level, long-term loss of work, and high Oswestry Disability Index (ODI) score at baseline have been identified as negative prognostic indicators.13 The associations of preoperative modifiable risk factors with clinical outcomes after TDR surgery are not well characterized.

The activL artificial disc is a next-generation TDR device that demonstrated improved efficacy and safety compared with control TDR discs (ProDisc-L or Charité) at 2- and 5-year follow-up evaluations in a U.S. Food and Drug Administration (FDA)–required Investigational Device Exemption (IDE) randomized controlled trial.3,14 The final study results with 7-year follow-up are now available.15 The objective of the current study was to conduct a post hoc analysis of prospectively collected data included in the 7-year activL IDE data set in order to develop multivariable prediction models for the identification of preoperative factors that are predictive of better long-term outcomes after lumbar TDR and that could be used to optimize preoperative patient care.

Methods

Patient Population

We analyzed data from the multicenter, IDE randomized controlled trial of activL (NCT00589797). Detailed IDE trial methods have been previously reported.3 Briefly, eligible patients reported lumbar pain due to a radiographically confirmed diagnosis of DDD at a single symptomatic level (L4–5 or L5–S1) after failure of nonsurgical management for ≥ 6 months. Patients with greater than grade 1 spondylolisthesis were ineligible. Patients were randomly allocated (2:1 ratio) to undergo implantation with the activL disc or a control artificial disc (ProDisc-L or Charité). Patients returned for follow-up visits at 6 weeks, 3 months, and 6 months, as well as annually thereafter for 7 years. Standardized patient-reported outcome measures and findings on physical examination, neurological assessment, and 6-view radiography were obtained at each follow-up visit.

The randomized patient set (n = 283) from the activL trial was used to identify candidate factors and develop the final models and risk equations. A nonrandomized training patient set (n = 52) was used to validate the models and risk equations. To be considered in the present analysis, patients must have received either the activL or ProDisc device. Patients who received the Charité device were excluded from the present analysis because this device is no longer commercially available and therefore not considered relevant to current patient care.

Preoperative Factors and Outcomes of Interest

We assessed the following preoperative factors included in the activL trial for their ability to predict postoperative outcomes over the 7-year follow-up period: age, sex, baseline BMI, work status and type of labor before and after sustaining back injury, recreation status before and after sustaining back injury, any pain medication use, narcotics use, history of previous lumbar surgery such as microdiscectomy, baseline disc angle, baseline disc height, index level, baseline flexion-extension range of motion (ROM), baseline translation, baseline ODI score, baseline visual analog scale (VAS)–back pain score, baseline VAS–leg pain score of the worst leg (i.e., the leg with the more severe VAS–leg pain score), duration of back pain, baseline quality of life as assessed with SF-36 physical component summary (PCS) score, and randomization to receive the activL or ProDisc device.

Key outcomes relevant to patients who undergo TDR were measured over the 7-year follow-up period and included VAS–back pain score, ODI score, ROM, SF-36 PCS score, reoperation, serious adverse events (SAEs), return to work, return to recreation, patient satisfaction level, and patient willingness to have surgery again. The continuous variables of VAS–back pain score, ODI score, ROM, and SF-36 PCS were dichotomized as equal to or better than the mean or worse than the mean of the IDE trial. All other outcomes were dichotomous events. For return to work and return to recreation, the analyses focused on identification of preoperative predictors of return to the previous activity level before back injury.

Statistical Analysis

Statistical methods and reporting followed the TRIPOD Checklist.16 Continuous data were reported as mean ± SD, and categorical data were reported as frequencies and percentages, unless otherwise noted.

Identification of Prognostic Factors

A two-step approach was used to identify preoperative factors that predicted each postoperative outcome in the randomized patient set. First, univariable regression models were used to identify candidate factors associated with the outcome. Logistic regression was used for all outcomes and results were presented as odds ratio (OR) with 95% CI, where OR > 1 represented a benefit for positive events and OR < 1 represented a benefit for negative events. Cox proportional hazards models were used to model time-to-event outcomes (i.e., time to return to work). Results were presented as hazard ratio (HR) with 95% CI. Because return to work was a positive event for patients, HR > 1 represented a benefit (i.e., time to return to work) for the patient compared with the reference. Factors with p < 0.10 were entered into the multivariable regression models to identify predictive factors.

Second, multivariate regression models were used to identify predictive factors. The included candidate factors had p < 0.10 in the univariable regression models. A backward stepwise approach was used to select the final multivariable model. This process involved the inclusion of all candidate factors in the initial model, and factors were removed one by one if prediction error increased. Prediction error was measured with the Akaike information criterion to compare different models and to determine which model fit best. The Akaike information criterion considers the number of estimated parameters in the model and how well the model reproduces data.17 Factors with p < 0.05 for the estimated ORs and HRs were considered statistically significant in the final multivariate regression model.

Risk Equations

Risk equations were developed using the final multivariable logistic regression models for better/worse than the 7-year mean VAS–back pain score, better/worse than the 7-year mean ODI score, reoperation within 7 years, and return to preinjury full-time work status within 7 years. The equations were used to predict the probability of patients achieving these outcomes on the basis of the preoperative factors included in the final models.

Model Validation

Internal validation of the multivariate regression models was performed using the original data set. External validation was performed independently after the risk equations were formulated by using the nonrandomized training patient set from the activL trial. A total of 55 patients were not randomly selected to receive a device, of whom 3 did not undergo device implantation; thus, the training set included 52 nonrandomized patients. The formulated risk equations were applied to these 52 patients to estimate the predicted probability of achieving good outcomes. The predictive power of the multivariate regression models and risk equations was assessed using area under the receiver operating characteristic curve (AUC) analysis. AUC ranges from 0 to 1, with values closer to 1 indicating better prediction. Calibration of the risk equations was evaluated using the pooled Wald test.18 In this study, p > 0.05 suggested no evidence of poor calibration.

Missing Data

Multiple imputation with predictive mean matching19 was used to impute values that were missing from the original data set (i.e., missing values were imputed into the original data set); multiply imputed and complete data sets were created from the original data set. Predictive mean matching introduces randomness to the imputed values in order to properly account for the inherent uncertainties of the missing values.20 Statistical analyses (e.g., logistic regression) were performed on each imputed data set. Parameters of interest (e.g., OR and HR) and their variance were calculated by pooling individual estimates across the models built with the imputed data sets (Appendix 1).

Results

Baseline Patient Characteristics and 7-Year Outcomes

Of the 396 patients included in the activL IDE clinical trial data set, 113 were excluded from the present analysis because they were nonrandomized training cases (n = 52), were randomized to the Charité control arm (n = 41), or withdrew prior to the procedure (n = 20). Therefore, a total of 283 patients were included in the present analysis. The baseline characteristics and 7-year outcomes of these patients are presented in Tables 1 and 2.

TABLE 1.

Baseline patient characteristics

CharacteristicPooled Value (n = 283)
Demographic
 Age39 ± 9
 Male149 (53)
 BMI27 ± 4
Medical history
 Current narcotics use242 (86)
 Musculoskeletal/connective tissue120 (42)
 Smoking history63 (22)
 Gastrointestinal85 (30)
 Cardiovascular86 (30)
 Neurologic75 (27)
 Previous lumbar surgery70 (25)
 Cervical pain55 (19)
 Pulmonary44 (16)
 Endocrine/metabolic31 (11)
 Renal29 (10)
 Hepatic/biliary23 (8)
Symptom score
 VAS–back pain79 ± 15
 ODI57 ± 14
SF-36 health-related quality-of-life score
 PCS30 ± 6
 Mental component summary40 ± 14
Radiographic
 Decreased disc height213 (75)
 Herniated disc199 (70)
 Facet joint degeneration72 (25)
 Facet joint osteophyte51 (18)
 LF, AF, or facet joint hypertrophy56 (20)
 Instability26 (9)
 Vacuum phenomenon20 (7)
ROM
 FE rotation, °5.4 (−1.4 to 26.9)
 FE translation, mm0.3 (−1.4 to 3.8)

AF = annulus fibrosus; FE = flexion-extension; LF = ligamentum flavum. Values are shown as mean ± SD, number (percent), or median (range).

TABLE 2.

Outcome results at 7 years

Categorical PredictorValue*
VAS–back pain score at 7 yrs18.63 ± 24.52 (0 to 90)
 Better than mean102 (65.8)
 Worse than mean53 (34.2)
ODI score at 7 yrs17.63 ± 18.05 (0 to 74)
 Better than mean93 (60.0)
 Worse than mean62 (40.0)
FE ROM5.11 ± 4.53 (−0.9 to 20.8)
 Better than mean58 (37.9)
 Worse than mean95 (62.1)
SF-36 PCS score47.72 ± 10.69 (21.8 to 63.6)
 Better than mean85 (55.2)
 Worse than mean69 (44.8)
Safety
 Device-related SAE
  Yes18 (6.4)
  No265 (93.6)
 Reoperation
  Yes14 (4.9)
  No265 (95.1)
Patient satisfaction
 Very satisfied134 (86.5)
 Less than very satisfied21 (13.5)
Willingness to have surgery again
 Definitely128 (82.6)
 Not definitely27 (17.4)
Return to work & recreation
 Full-time w/o restrictions98 (68.5)
 Other45 (31.5)
Work type
 Manual labor58 (48.3)
 Sedentary62 (51.7)
Recreation type
 Contact or noncontact sports101 (65.2)
 Sedentary54 (34.8)
Time to return to work, mos3.25 (2.27 to 4.30)
Returned to work115 (75.7)

Values are shown as mean ± SD (range), number (percent), or median (95% CI).

Results at 7 years were based on complete case analysis (i.e., not multiply imputed).

Preoperative Predictors of Achieving Better Than the Mean Scores in the ActivL Trial

Overall, predictive factors were identified for achieving better than the mean VAS–back pain and ODI scores. No predictive factors were identified for improved ROM and SF-36 PCS score (Table 3).

TABLE 3.

Preoperative predictors of achieving better than the mean score at 7 years in the activL IDE trial

FactorUnivariate Regression ModelMultivariate Regression ModelAUC (95% CI)HL Test (p Value)
OR (95% CI)*p ValueOR (95% CI)*p Value
VAS–back pain score
 Preop narcotics use0.51 (0.28–0.94)0.0380.56 (0.30–1.05)0.0750.680 (0.596–0.754)>0.9999
 Preop mean disc height (per unit decrease)1.19 (1.01–1.39)0.0421.15 (0.98–1.35)0.101
 Preop VAS–back pain score (per unit increase)0.97 (0.95–1.00)0.0390.97 (0.95–1.00)0.056
 Randomization to receive activL device (vs ProDisc)2.06 (1.11–3.82)0.0242.20 (1.13–4.30)0.024
ODI score
 Preop narcotics use0.48 (0.28–0.81)0.0070.47 (0.26–0.84)0.0130.731 (0.668–0.785)>0.9999
 Preop mean disc height (per unit decrease)1.19 (1.00–1.41)0.0721.16 (0.95–1.43)0.159
 Preop VAS–back pain score (per unit increase)0.97 (0.95–0.99)0.0070.98 (0.96–1.00)0.07
 Preop VAS–leg pain score of worst leg (per unit increase)0.99 (0.98–1.00)0.0170.99 (0.98–1.00)0.058
 Randomization to receive activL device (vs ProDisc)3.72 (1.72–8.07)0.0034.48 (1.89–10.63)0.003
FE ROM score§
 BMI (per unit increase)1.06 (0.99–1.13)0.0821.06 (0.99–1.13)0.0820.562 (0.517–0.607)0.9123
SF-36 PCS score
 Preop narcotics use0.55 (0.28–1.06)0.0910.55 (0.28–1.06)0.0910.568 (0.511–0.624)>0.9999

HL = Hosmer-Lemeshow. Boldface type indicates statistical significance.

OR < 1 indicates decreased probability of having better than the mean score; OR > 1 indicates greater probability of having better than the mean score.

The mean score was 19.

The mean score was 18.

The mean score was 5.

The mean score was 48.

For VAS–back pain, the univariate regression model identified 4 candidate factors (preoperative narcotics use, decreasing preoperative disc height, greater preoperative VAS–back pain score, and randomization to receive activL device). In the multivariate model, randomization to receive the activL device was significantly associated with a 120% increase in the odds of achieving a better VAS–back pain score than the mean at 7 years, as compared with patients who were randomized to receive the ProDisc-L device (OR 2.20, 95% CI 1.13–4.30, p = 0.024). The prediction accuracy of the multivariable model was poor, with AUC of 0.680.

For ODI score, the univariate regression model identified 5 candidate factors (preoperative narcotics use, decreasing preoperative disc height, greater preoperative VAS–back pain score, greater preoperative VAS–leg pain score of the worst leg, and randomization to receive the activL disc). In the multivariate model, preoperative narcotics use and randomization to receive the activL disc were significantly associated with achieving a better ODI score than the mean at 7 years. Preoperative narcotics use was associated with a 53% reduction in the odds of achieving better ODI score than the mean at 7 years compared with no preoperative narcotics use (OR 0.47, 95% CI 0.26–0.84, p = 0.013). Patients who were randomized to receive the activL device had a 348% increase in the odds of achieving a better ODI score than the mean at 7 years when compared with patients who randomly received the ProDisc-L device (OR 4.48, 95% CI 1.89–10.63, p = 0.003). The prediction accuracy of the multivariable model was reasonable, with AUC of 0.731.

For ROM, the univariate regression model identified greater preoperative BMI as a candidate factor; however, the multivariate regression model did not identify a statistically significant association between this factor and mean ROM of 5° or greater (OR 1.06, 95% CI 0.99–1.13, p = 0.082). The prediction accuracy of the multivariable model was poor, with AUC of 0.562.

Similarly, for SF-36 PCS score, only preoperative narcotics use was identified as a candidate factor in the univariate regression model. However, this association was not statistically significant in the multivariate regression model (OR 0.55, 95% CI 0.28–1.06, p = 0.091). The prediction accuracy of the multivariable model was poor, with AUC of 0.568.

Similarly, for SF-36 PCS score, only preoperative narcotics use was identified as a candidate factor in the univariate regression model. However, this association was not statistically significant in the multivariate regression model (OR 0.55, 95% CI 0.28–1.06, p = 0.091). The prediction accuracy of the multivariable model was poor, with AUC of 0.568.

Preoperative Predictors of Outcomes Related to Safety, Return to Work, and Patient Satisfaction

Multiple preoperative factors were associated with the additional outcomes of SAE, reoperation, return to full-time work without restrictions, return to manual labor, return to recreation, patient satisfaction, and patient willingness to have surgery again (Fig. 1, Appendix 2). Significant predictive factors included younger age (for freedom from SAEs, OR 0.93; presence of SAEs associated with older age, OR 1.07, 95% CI 1.01–1.14, p = 0.026), greater preoperative VAS–leg pain (reoperation, OR 1.03, 95% CI 1.00–1.05, p = 0.024), female sex (not returning to unrestricted full-time work, OR 0.20, 95% CI 0.11–0.37, p < 0.001; slower return to work, HR 0.64, 95% CI 0.48–0.84, p = 0.002), preoperative narcotics use (not returning to unrestricted full-time work, OR 0.42, 95% CI 0.22–0.80, p = 0.01), working manual labor after back injury but before lumbar TDR (returning to manual labor work, OR 2.79, 95% CI 1.24–6.28, p = 0.023), unrestricted full-time work before back injury (quicker return to work, HR 1.85, 95% CI 1.07–3.20, p = 0.029), unrestricted full-time work after back injury and before lumbar TDR (quicker return to work, HR 1.64, 95% CI 1.21–2.23, p = 0.001), working manual labor before back injury (slower return to work, HR 0.58, 95% CI 0.43–0.78, p < 0.001), greater preoperative ODI score (slower return to work, HR 0.98, 95% CI 0.97–0.99, p < 0.001), decreasing preoperative disc height (patient willingness to have surgery again, OR 1.43, 95% CI 1.07–1.89, p = 0.031), and randomization to receive the activL device (patient willingness to have surgery again, OR 5.29, 95% CI 2.15–13.01, p = 0.002). No factors were significantly associated with return to recreation and patient satisfaction. Overall, prediction accuracy of the multivariable models was poor or reasonable.

FIG. 1.
FIG. 1.

Summary of predictive factors for safety, return to work, and patient satisfaction outcomes at 7-year follow-up. + = positive predictor of event; − = negative predictor of event.

Validation

The risk equations for probability of achieving better VAS–back pain and ODI scores than the 7-year mean values, reoperation, and return to unrestricted full-time work after lumbar TDR (Table 4) were externally validated using nonrandomized patients from the activL trial. These patients had similar baseline characteristics and 7-year outcomes as the randomized patients (Appendix 3).

TABLE 4.

Summary of risk equations

OutcomeRisk Equation
Probability of achieving better VAS–back pain score than the 7-yr mean (i.e., 19)Probability of achieving better VAS–back pain score = 1 / (1 + e-[3.66 + X]) Where X = −0.14 × preop mean disc height + 0.57 × [preop narcotics use = yes] + 0.79 × [randomized to activL = yes] − 0.03 × preop VAS–back pain score*
Probability of achieving better ODI score than the 7-yr mean (i.e., 18)Probability of achieving better ODI score = 1 / (1 + e-[3.00 + X]) Where X = −0.75 × [preop narcotics use = yes] + 1.50 × [randomized to activL = yes] − 0.02 × preop VAS–back pain score − 0.02 × preop VAS–leg pain score − 0.16 × preop mean disc height*
Probability of reoperationProbability of reoperation = 1 / (1 + e-[−5.90 + X]) Where X = 2.05 × [preop narcotics use = yes] + 0.03 × preop VAS–leg pain score*
Probability of returning to full-time work w/o restrictionsProbability of returning to full-time work w/o restrictions = 1 / (1 + e-[0.19 + X]) Where X = −1.60 × [sex = female] − 0.87 × [preop narcotics use = yes] + 0.06 × preop SF-36 PCS score + 0.77 × [work type after injury = manual labor]*

“Yes” has a value of 1 and “No” has a value of 0 for preoperative narcotics use and randomization to receive activL device.

“Sex” has a value of 1 for female and 0 for male.

“Work type after injury” has a value of 1 for manual labor and 0 for sedentary work.

Overall, the risk equations had strong prediction accuracy when the predicted probabilities for each outcome were compared with the results observed in the activL trial (Table 5). For example, 33% (17/52) of patients were predicted to have > 70% probability of achieving a VAS–back pain score ≤ 19 at 7 years. Results from the activL trial showed that 71% (12/17) of these patients achieved scores better than the mean 7-year VAS–back pain score. Prediction accuracy based on AUC ranged from poor to good.

TABLE 5.

Validation results of the risk equations

Estimated ProbabilityVAS Score <19ODI Score <18ReoperationReturned to Working Full-Time w/o Restrictions
PredictedObservedPredictedObservedPredictedObservedPredictedObserved
0<P≤0.1000039/52 (75)0/39 (0)00
0.1<P≤0.2002/52 (4)1/2 (50)11/52 (21)0/11 (0)1/52 (2)0/1 (0)
0.2<P≤0.3001/52 (2)02/52 (4)0/2 (0)4/52 (8)1/4 (25)
0.3<P≤0.42/52 (4)1/2 (50)10/52 (19)2/10 (20)009/52 (17)6/9 (67)
0.4<P≤0.511/52 (21)4/11 (36)7/52 (13)2/7 (29)006/52 (12)5/6 (83)
0.5<P≤0.68/52 (15)4/8 (50)6/52 (12)3/6 (50)002/52 (4)2/2 (100)
0.6<P≤0.714/52 (27)8/14 (57)11/52 (21)7/11 (64)0011/52 (21)11/11 (100)
0.7<P≤0.811/52 (21)7/11 (64)8/52 (15)4/8 (50)009/52 (17)8/9 (89)
0.8<P≤0.96/52 (12)5/6 (83)7/52 (13)5/7 (71)007/52 (13)7/7 (100)
0.9<P≤10000003/52 (6)3/3 (100)
AUC (95% CI)0.605 (0.465–0.729)0.656 (0.511–0.777)NE*0.801 (0.529–0.935)
p value0.99970.95780.90790.2123

NE = not estimated.

Values are shown as number (percent) unless indicated otherwise.

No patients in the training set underwent reoperation within 7 years.

Determined with the HL test.

Discussion

Although the ideal indications for lumbar TDR, as elucidated in multiple IDE studies required by the U.S. FDA, have been well characterized, patient selection remains a diagnostic challenge. Patients who present with axial, mechanical low-back pain are generally an inherently heterogeneous population, and surgeons would benefit from insight into positive predictive characteristics. The specific patient population studied in this activL cohort analysis is especially complex, with high rates of preoperative pain and disability.

This combination of heterogeneity with substantial morbidity makes proper preoperative patient selection even more crucial to achieve positive patient outcome. The scope of this analysis was to determine whether a specific subset of patients from the IDE study had even better outcomes than the overall positive study outcomes. Our results identified multiple factors associated with improved outcomes, such as no preoperative narcotics use, unrestricted full-time work, working manual labor before and after back injury, working a sedentary job before back injury, and randomization to receive the activL device. Preoperative BMI, VAS–back pain score, ODI score, back pain duration ≥ 1 year, SF-36 PCS score, and recreational activity were not associated with outcomes.

A meta-analysis of four studies that compared TDR to fusion in patients with lumbar DDD showed that TDR was associated with a statistically significantly greater chance of improvement in ODI and VAS–back pain scores, lower risk of reoperation, and improved patient satisfaction at 5-year follow-up.21 Notably, the current analysis did not identify patients eligible for treatment; rather, we suggest factors to help identify patients who may achieve a better-than-average response with TDR. Surgeons would then have the opportunity to tailor preoperative care in order to optimize outcomes for these patients. For example, our analysis estimated that, assuming other factors in the model were equal, patients who used narcotics preoperatively had a 53% reduction in the odds, or a 15% decrease in the probability, of achieving an ODI score of 18 or better.

In this analysis, preoperative narcotics use was associated with greater ODI score and decreased likelihood of return to unrestricted full-time work within 7 years. Although studies in the literature on the relationship between preoperative narcotics use and long-term lumbar TDR outcomes are limited, multiple studies of lumbar spine surgery, including single-level lumbar fusion, have associated preoperative narcotics use with greater pain and disability, decreased likelihood of return to work, and revision surgery.22–26 Our analysis also revealed a trend for increased likelihood of reoperation within 7 years after TDR in patients with preoperative narcotics use, although this association was not significant (p = 0.051).

Preoperative work status was also associated with improved long-term TDR outcomes in our study. Specifically, working full-time either before or after injury, but before TDR, was associated with quicker return to work without restrictions. Patients were also more likely to return to work if they were working manual labor, even if they were working after back injury. However, patients returned to work more quickly if working a sedentary job before back injury. These findings contradict those from studies that retrospectively analyzed data from the National Neurosurgery Quality and Outcomes Database (QOD, formerly N2QOD) and reported decreased likelihood of return to work after surgery among patients with physically demanding jobs at the time surgery.27,28

We found that randomization to receive the activL device may have further improved the odds of achieving better than the mean VAS–back pain score by 120% and ODI score by 348%, as well as patient willingness to have surgery again by 429%, when compared with patients who randomly received the ProDisc-L device. Short- and long-term analyses of the activL study are in accordance with these findings.3,14,15 Intuitively, factors associated with particular TDR devices, such as ease of implantation and unique biomechanical characteristics, should have implications on clinical outcomes. In this analysis, surgeon choice to use the activL device was associated with improved outcomes.

We also identified greater VAS–leg pain score as a strong risk factor for reoperation. Of the 13 reoperations in the activL trial, 5 patients (3 underwent decompression foraminotomy owing to postoperative leg pain, and 2 underwent lumbar fusion owing to postoperative low-back and leg pain) had leg pain that persisted from baseline. Radiculopathy and VAS–leg pain score ≥ 40 were associated with poor patient outcomes,29 suggesting that the ideal patients for TDR are those with predominately mechanical low-back pain. Therefore, a significant subset of patients who present primarily with chronic radiculopathy may have neuropathic pain that would remain recalcitrant to any surgical intervention. No factors were associated with long-term improvement in ROM, SF-36 PCS score, return to recreation, and patient satisfaction after lumbar TDR.

The risk equations derived from the final models included in the current analysis provide a method for estimating the likelihood that a patient achieves an outcome at the 7-year follow-up after TDR, and these equations could be developed into a tool that surgeons could use to optimize preoperative management strategies and further improve outcomes. In an effort to facilitate decision-making activities and manage patient expectations, similar tools have been published for predicting outcomes of lumbar spine surgical procedures for DDD, primary stenosis, spondylolisthesis, disc herniation, and symptomatic mechanical disc collapse.30–35 To our knowledge, there is no such tool for lumbar TDR patients. Systematic literature reviews report limited evidence for risk prediction in lumbar spine surgical procedures and highlight it as an emerging field and area for growth.36,37 The extent to which risk prediction tools are used in clinical practice for lumbar spine surgical procedures is not well known. Understanding the adoption of these tools and how well they predict patient outcomes in real-world practice will also be vital to managing patient expectations and optimizing patient care prior to surgery. These tools may improve patient selection and preoperative counseling for patients considering TDR.

Our study had several strengths. Predictive modeling was based on prospectively collected data from the activL trial. Data from the activL and ProDisc-L devices were pooled to maximize sample size and to model predictions for TDR as a class. Additionally, the sample size was preserved with multiple imputations of missing values. This allowed investigation of the relationships between multiple preoperative factors and several key clinical endpoints, including disability, back pain, adverse events, reoperation, return to work, and return to recreation. Additional model validations leveraged external data that were not used to develop the models used to assess prediction accuracy and model fit.

This study was a post hoc analysis of a subset of patients from a randomized controlled trial and is therefore subject to the inherent statistical limitations associated with such an analysis. These predictive models may not be fully generalizable to real-world contexts. Clinical trial data were based on selective inclusion and exclusion criteria, and the patients included in the activL trial had substantial morbidity. The preoperative mean VAS–back pain and ODI scores were high, as were the proportions of patients who received narcotics at baseline, patients who had history of lumbar surgery, and patients who were smokers. In clinical practice, however, patients who undergo TDR may be healthier than those selected for the clinical trial. Additionally, the AUC values used to assess the predictive power of the regression models ranged from 0.562 (poor association) to 0.772 (reasonable association). Analyses were also limited by the data and preoperative factors collected in the activL trial. It is unclear how long patients need to be free from narcotics preoperatively before an improvement in outcomes after lumbar arthroplasty would be seen because these data were unavailable in the activL data set. Other potentially notable factors that were also unavailable included facet disease, spondylolisthesis, and healthcare insurance (workers’ compensation vs private insurance).

Conclusions

Lumbar TDR is a well-established treatment option for symptomatic lumbar DDD and has good long-term patient outcomes. Tailoring patient care before TDR may further improve patient outcomes. Discontinuation of narcotics may further improve patient outcomes, as our analysis identified associations between lack of preoperative narcotics use and further improvement in ODI score relative to the mean score in the activL IDE trial and increased likelihood of return to work within 7 years. Other preoperative factors that may further improve outcomes included working full-time without restrictions, working manual labor, working a sedentary job before back injury, and randomization to receive the activL device. A future goal may be to develop a user-friendly tool for clinicians and patients based on the risk equations developed for prediction of VAS–back pain score, ODI score, reoperation, and return to work in order to assist clinicians with preoperative care optimization for select patients undergoing TDR.

Acknowledgments

We thank Chris Cameron of CRG-EVERSANA for statistical expertise and Katie Kleinschuster, Tabitha Pitten, and Andrea Vovk of Aesculap, Center Valley, Pennsylvania, for constructive discussions regarding the study design.

Disclosures

The devices that are the subject of this article were evaluated as part of a U.S. FDA–approved investigational protocol (IDE) or corresponding national protocol for the treatment of single-level DDD of the lumbar spine (L4–S1) in patients who have not responded to at least 6 months of conservative care. Grant funds from Aesculap Implant Systems, LLC, Center Valley, Pennsylvania, were received in support of this work. Aesculap Implant Systems, LLC, was involved in the design and conduct of the study, management of data, and approval of the manuscript. Dr. Coric is a consultant for Aesculap, Globus Medical, and Medtronic; and receives royalties from RTI Surgical. Dr. Zigler is a consultant for Aesculap. Dr. Derman is a consultant for Neo Spine, Orthofix, Degen Medical, and Integrity Implants; receives royalties from Degen Medical; receives research support from Orthofix; is on the speakers bureau of Joimax and Integrity Implants; receives clinical or research support for the study described from Aesculap; and received statistical analysis or writing or editorial assistance for this study from Aesculap. Dr. Braxton is a consultant for Aesculap. Mr. Situ is an employee of EVERSANA. Dr. Patel is an employee of EVERSANA Life Science Services LLC.

Author Contributions

Conception and design: all authors. Analysis and interpretation of data: all authors. Drafting the article: Patel. Critically revising the article: Patel, Coric, Zigler, Derman, Braxton. Reviewed submitted version of manuscript: Patel, Coric, Zigler, Derman, Braxton. Approved the final version of the manuscript on behalf of all authors: Patel. Statistical analysis: Situ. Administrative/technical/material support: Patel.

Supplemental Information

Online-Only Content

Supplemental material is available with the online version of the article.

References

  • 1

    Geurts JW, Willems PC, Kallewaard JW, et al. The impact of chronic discogenic low back pain: costs and patients’ burden. Pain Res Manag. 2018;2018:4696180.

  • 2

    Crow WT, Willis DR. Estimating cost of care for patients with acute low back pain: a retrospective review of patient records. J Am Osteopath Assoc. 2009;109(4):229233.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 3

    Garcia R Jr, Yue JJ, Blumenthal S, et al. Lumbar total disc replacement for discogenic low back pain: two-year outcomes of the activL multicenter randomized controlled IDE clinical trial. Spine (Phila Pa 1976).2015;40(24):18731881.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 4

    Gornet M, Dryer R, Peloza J, Schranck F. Lumbar disc arthroplasty vs. Anterior lumbar interbody fusion: Five-year outcomes for patients in the Maverick® disc IDE study. Spine J. 2010;10(9)(suppl):S64.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 5

    Gornet MF, Burkus JK, Dryer RF, Peloza JH. Lumbar disc arthroplasty with Maverick disc versus stand-alone interbody fusion: a prospective, randomized, controlled, multicenter investigational device exemption trial. Spine (Phila Pa 1976).2011;36(25):E1600E1611.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 6

    Guyer RD, McAfee PC, Banco RJ, et al. Prospective, randomized, multicenter Food and Drug Administration investigational device exemption study of lumbar total disc replacement with the CHARITE artificial disc versus lumbar fusion: five-year follow-up. Spine J. 2009;9(5):374386.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 7

    Guyer RD, Pettine K, Roh JS, et al. Five-year follow-up of a prospective, randomized trial comparing two lumbar total disc replacements. Spine (Phila Pa 1976).2016;41(1):38.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 8

    Zigler J, Delamarter R, Spivak JM, et al. Results of the prospective, randomized, multicenter Food and Drug Administration investigational device exemption study of the ProDisc-L total disc replacement versus circumferential fusion for the treatment of 1-level degenerative disc disease. Spine (Phila Pa 1976).2007;32(11):11551163.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 9

    Blumenthal S, McAfee PC, Guyer RD, et al. A prospective, randomized, multicenter Food and Drug Administration investigational device exemptions study of lumbar total disc replacement with the CHARITE artificial disc versus lumbar fusion: part I: evaluation of clinical outcomes. Spine (Phila Pa 1976).2005;30(14):15651575,E387E391.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 10

    Hellum C, Johnsen LG, Storheim K, et al. Surgery with disc prosthesis versus rehabilitation in patients with low back pain and degenerative disc: two year follow-up of randomised study. BMJ. 2011;342:d2786.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 11

    Guyer RD, Siddiqui S, Zigler JE, et al. Lumbar spinal arthroplasty: analysis of one center’s twenty best and twenty worst clinical outcomes. Spine (Phila Pa 1976).2008;33(23):25662569.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 12

    Gornet MF, Schranck F, Wharton ND, et al. Optimizing success with lumbar disc arthroplasty. Eur Spine J. 2014;23(10):21272135.

  • 13

    Furunes H, Hellum C, Brox JI, et al. Lumbar total disc replacement: predictors for long-term outcome. Eur Spine J. 2018;27(3):709718.

  • 14

    Yue JJ, Garcia R, Blumenthal S, et al. Five-year results of a randomized controlled trial for lumbar artificial discs in single-level degenerative disc disease. Spine (Phila Pa 1976).2019;44(24):16851696.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 15

    Radcliff K, Zigler J, Braxton E, et al. Final long-term reporting from a randomized controlled IDE trial for lumbar artificial discs in single-level degenerative disc disease: 7-year results. Int J Spine Surg. 2021;15(4):612632.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 16

    Collins GS, Reitsma JB, Altman DG, Moons KG. Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): the TRIPOD statement. Ann Intern Med. 2015;162(1):5563.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 17

    Akaike H. A new look at the statistical model identification. IEEE Trans Automat Contr. 1974;19(6):716723.

  • 18

    Li KH, Meng XL, Raghunathan TE, Rubin DB. Significance levels from repeated p-values with multiply-imputed data. Stat Sin. 1991;1:6592.

    • Search Google Scholar
    • Export Citation
  • 19

    Van Buuren S. Flexible Imputation of Missing Data. CRC Press;2018.

  • 20

    Marshall A, Altman DG, Holder RL, Royston P. Combining estimates of interest in prognostic modelling studies after multiple imputation: current practice and guidelines. BMC Med Res Methodol. 2009;9(1):57.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 21

    Zigler J, Gornet MF, Ferko N, et al. Comparison of lumbar total disc replacement with surgical spinal fusion for the treatment of single-level degenerative disc disease: a meta-analysis of 5-year outcomes from randomized controlled trials. Global Spine J. 2018;8(4):413423.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 22

    Kim S, Ozpinar A, Agarwal N, et al. Relationship between preoperative opioid use and postoperative pain in patients undergoing minimally invasive stand-alone lateral lumbar interbody fusion. Neurosurgery. 2020;87(6):11671173.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 23

    Yerneni K, Nichols N, Abecassis ZA, et al. Preoperative opioid use and clinical outcomes in spine surgery: a systematic review. Neurosurgery. 2020;86(6):E490E507.

  • 24

    Bhattacharjee S, Pirkle S, Shi LL, Lee MJ. The effects of chronic preoperative opioid use on single-level lumbar fusion outcomes. Clin Spine Surg. 2020;33(8):E401E406.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 25

    Zakaria HM, Mansour TR, Telemi E, et al. The association of preoperative opioid usage with patient-reported outcomes, adverse events, and return to work after lumbar fusion: analysis from the Michigan Spine Surgery Improvement Collaborative (MSSIC). Neurosurgery. 2020;87(1):142149.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 26

    Nguyen TH, Randolph DC, Talmage J, et al. Long-term outcomes of lumbar fusion among workers’ compensation subjects: a historical cohort study. Spine (Phila Pa 1976).2011;36(4):320331.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 27

    Khan I, Bydon M, Archer KR, et al. Impact of occupational characteristics on return to work for employed patients after elective lumbar spine surgery. Spine J. 2019;19(12):19691976.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 28

    Asher AL, Devin CJ, Archer KR, et al. An analysis from the Quality Outcomes Database, Part 2. Predictive model for return to work after elective surgery for lumbar degenerative disease. J Neurosurg Spine. 2017;27(4):370381.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 29

    Zweig T, Aghayev E, Melloh M, et al. Influence of preoperative leg pain and radiculopathy on outcomes in mono-segmental lumbar total disc replacement: results from a nationwide registry. Eur Spine J. 2012;21(6 Suppl):S729S736.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 30

    Alentado VJ, Caldwell S, Gould HP, et al. Independent predictors of a clinically significant improvement after lumbar fusion surgery. Spine J. 2017;17(2):236243.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 31

    Lee MJ, Cizik AM, Hamilton D, Chapman JR. Predicting medical complications after spine surgery: a validated model using a prospective surgical registry. Spine J. 2014;14(2):291299.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 32

    McGirt MJ, Sivaganesan A, Asher AL, Devin CJ. Prediction model for outcome after low-back surgery: individualized likelihood of complication, hospital readmission, return to work, and 12-month improvement in functional disability. Neurosurg Focus. 2015;39(6):E13.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 33

    Veeravagu A, Li A, Swinney C, et al. Predicting complication risk in spine surgery: a prospective analysis of a novel risk assessment tool. J Neurosurg Spine. 2017;27(1):8191.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 34

    McGirt MJ, Bydon M, Archer KR, et al. An analysis from the Quality Outcomes Database, Part 1. Disability, quality of life, and pain outcomes following lumbar spine surgery: predicting likely individual patient outcomes for shared decision-making. J Neurosurg Spine. 2017;27(4):357369.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 35

    Khor S, Lavallee D, Cizik AM, et al. Development and validation of a prediction model for pain and functional outcomes after lumbar spine surgery. JAMA Surg. 2018;153(7):634642.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 36

    White HJ, Bradley J, Hadgis N, et al. Predicting patient-centered outcomes from spine surgery using risk assessment tools: a systematic review. Curr Rev Musculoskelet Med. 2020;13(3):247263.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 37

    Osorio JA, Scheer JK, Ames CP. Predictive modeling of complications. Curr Rev Musculoskelet Med. 2016;9(3):333337.

Supplementary Materials

  • View in gallery

    Summary of predictive factors for safety, return to work, and patient satisfaction outcomes at 7-year follow-up. + = positive predictor of event; − = negative predictor of event.

  • 1

    Geurts JW, Willems PC, Kallewaard JW, et al. The impact of chronic discogenic low back pain: costs and patients’ burden. Pain Res Manag. 2018;2018:4696180.

  • 2

    Crow WT, Willis DR. Estimating cost of care for patients with acute low back pain: a retrospective review of patient records. J Am Osteopath Assoc. 2009;109(4):229233.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 3

    Garcia R Jr, Yue JJ, Blumenthal S, et al. Lumbar total disc replacement for discogenic low back pain: two-year outcomes of the activL multicenter randomized controlled IDE clinical trial. Spine (Phila Pa 1976).2015;40(24):18731881.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 4

    Gornet M, Dryer R, Peloza J, Schranck F. Lumbar disc arthroplasty vs. Anterior lumbar interbody fusion: Five-year outcomes for patients in the Maverick® disc IDE study. Spine J. 2010;10(9)(suppl):S64.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 5

    Gornet MF, Burkus JK, Dryer RF, Peloza JH. Lumbar disc arthroplasty with Maverick disc versus stand-alone interbody fusion: a prospective, randomized, controlled, multicenter investigational device exemption trial. Spine (Phila Pa 1976).2011;36(25):E1600E1611.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 6

    Guyer RD, McAfee PC, Banco RJ, et al. Prospective, randomized, multicenter Food and Drug Administration investigational device exemption study of lumbar total disc replacement with the CHARITE artificial disc versus lumbar fusion: five-year follow-up. Spine J. 2009;9(5):374386.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 7

    Guyer RD, Pettine K, Roh JS, et al. Five-year follow-up of a prospective, randomized trial comparing two lumbar total disc replacements. Spine (Phila Pa 1976).2016;41(1):38.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 8

    Zigler J, Delamarter R, Spivak JM, et al. Results of the prospective, randomized, multicenter Food and Drug Administration investigational device exemption study of the ProDisc-L total disc replacement versus circumferential fusion for the treatment of 1-level degenerative disc disease. Spine (Phila Pa 1976).2007;32(11):11551163.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 9

    Blumenthal S, McAfee PC, Guyer RD, et al. A prospective, randomized, multicenter Food and Drug Administration investigational device exemptions study of lumbar total disc replacement with the CHARITE artificial disc versus lumbar fusion: part I: evaluation of clinical outcomes. Spine (Phila Pa 1976).2005;30(14):15651575,E387E391.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 10

    Hellum C, Johnsen LG, Storheim K, et al. Surgery with disc prosthesis versus rehabilitation in patients with low back pain and degenerative disc: two year follow-up of randomised study. BMJ. 2011;342:d2786.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 11

    Guyer RD, Siddiqui S, Zigler JE, et al. Lumbar spinal arthroplasty: analysis of one center’s twenty best and twenty worst clinical outcomes. Spine (Phila Pa 1976).2008;33(23):25662569.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 12

    Gornet MF, Schranck F, Wharton ND, et al. Optimizing success with lumbar disc arthroplasty. Eur Spine J. 2014;23(10):21272135.

  • 13

    Furunes H, Hellum C, Brox JI, et al. Lumbar total disc replacement: predictors for long-term outcome. Eur Spine J. 2018;27(3):709718.

  • 14

    Yue JJ, Garcia R, Blumenthal S, et al. Five-year results of a randomized controlled trial for lumbar artificial discs in single-level degenerative disc disease. Spine (Phila Pa 1976).2019;44(24):16851696.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 15

    Radcliff K, Zigler J, Braxton E, et al. Final long-term reporting from a randomized controlled IDE trial for lumbar artificial discs in single-level degenerative disc disease: 7-year results. Int J Spine Surg. 2021;15(4):612632.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 16

    Collins GS, Reitsma JB, Altman DG, Moons KG. Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): the TRIPOD statement. Ann Intern Med. 2015;162(1):5563.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 17

    Akaike H. A new look at the statistical model identification. IEEE Trans Automat Contr. 1974;19(6):716723.

  • 18

    Li KH, Meng XL, Raghunathan TE, Rubin DB. Significance levels from repeated p-values with multiply-imputed data. Stat Sin. 1991;1:6592.

    • Search Google Scholar
    • Export Citation
  • 19

    Van Buuren S. Flexible Imputation of Missing Data. CRC Press;2018.

  • 20

    Marshall A, Altman DG, Holder RL, Royston P. Combining estimates of interest in prognostic modelling studies after multiple imputation: current practice and guidelines. BMC Med Res Methodol. 2009;9(1):57.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 21

    Zigler J, Gornet MF, Ferko N, et al. Comparison of lumbar total disc replacement with surgical spinal fusion for the treatment of single-level degenerative disc disease: a meta-analysis of 5-year outcomes from randomized controlled trials. Global Spine J. 2018;8(4):413423.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 22

    Kim S, Ozpinar A, Agarwal N, et al. Relationship between preoperative opioid use and postoperative pain in patients undergoing minimally invasive stand-alone lateral lumbar interbody fusion. Neurosurgery. 2020;87(6):11671173.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 23

    Yerneni K, Nichols N, Abecassis ZA, et al. Preoperative opioid use and clinical outcomes in spine surgery: a systematic review. Neurosurgery. 2020;86(6):E490E507.

  • 24

    Bhattacharjee S, Pirkle S, Shi LL, Lee MJ. The effects of chronic preoperative opioid use on single-level lumbar fusion outcomes. Clin Spine Surg. 2020;33(8):E401E406.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 25

    Zakaria HM, Mansour TR, Telemi E, et al. The association of preoperative opioid usage with patient-reported outcomes, adverse events, and return to work after lumbar fusion: analysis from the Michigan Spine Surgery Improvement Collaborative (MSSIC). Neurosurgery. 2020;87(1):142149.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 26

    Nguyen TH, Randolph DC, Talmage J, et al. Long-term outcomes of lumbar fusion among workers’ compensation subjects: a historical cohort study. Spine (Phila Pa 1976).2011;36(4):320331.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 27

    Khan I, Bydon M, Archer KR, et al. Impact of occupational characteristics on return to work for employed patients after elective lumbar spine surgery. Spine J. 2019;19(12):19691976.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 28

    Asher AL, Devin CJ, Archer KR, et al. An analysis from the Quality Outcomes Database, Part 2. Predictive model for return to work after elective surgery for lumbar degenerative disease. J Neurosurg Spine. 2017;27(4):370381.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 29

    Zweig T, Aghayev E, Melloh M, et al. Influence of preoperative leg pain and radiculopathy on outcomes in mono-segmental lumbar total disc replacement: results from a nationwide registry. Eur Spine J. 2012;21(6 Suppl):S729S736.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 30

    Alentado VJ, Caldwell S, Gould HP, et al. Independent predictors of a clinically significant improvement after lumbar fusion surgery. Spine J. 2017;17(2):236243.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 31

    Lee MJ, Cizik AM, Hamilton D, Chapman JR. Predicting medical complications after spine surgery: a validated model using a prospective surgical registry. Spine J. 2014;14(2):291299.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 32

    McGirt MJ, Sivaganesan A, Asher AL, Devin CJ. Prediction model for outcome after low-back surgery: individualized likelihood of complication, hospital readmission, return to work, and 12-month improvement in functional disability. Neurosurg Focus. 2015;39(6):E13.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 33

    Veeravagu A, Li A, Swinney C, et al. Predicting complication risk in spine surgery: a prospective analysis of a novel risk assessment tool. J Neurosurg Spine. 2017;27(1):8191.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 34

    McGirt MJ, Bydon M, Archer KR, et al. An analysis from the Quality Outcomes Database, Part 1. Disability, quality of life, and pain outcomes following lumbar spine surgery: predicting likely individual patient outcomes for shared decision-making. J Neurosurg Spine. 2017;27(4):357369.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 35

    Khor S, Lavallee D, Cizik AM, et al. Development and validation of a prediction model for pain and functional outcomes after lumbar spine surgery. JAMA Surg. 2018;153(7):634642.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 36

    White HJ, Bradley J, Hadgis N, et al. Predicting patient-centered outcomes from spine surgery using risk assessment tools: a systematic review. Curr Rev Musculoskelet Med. 2020;13(3):247263.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 37

    Osorio JA, Scheer JK, Ames CP. Predictive modeling of complications. Curr Rev Musculoskelet Med. 2016;9(3):333337.

Metrics

All Time Past Year Past 30 Days
Abstract Views 431 431 0
Full Text Views 184 184 61
PDF Downloads 154 154 42
EPUB Downloads 0 0 0