Lumbar disc herniation is the most common cause of sciatica worldwide.1 Initial treatment of sciatica is conservative; however, approximately 20% of conservatively treated patients have intractable pain or disability that warrant consideration for discectomy.2–4
Approximately 500,000 lumbar discectomy procedures are performed annually in the United States alone.5 A major complication of this procedure is recurrent symptomatic disc herniation, which has been reported to occur in 7% to 18% of patients within 2 years postoperatively.6–8 Compared to the index event, recurrent symptomatic herniation is associated with worse clinical outcomes, surgical complications (i.e., cerebrospinal fluid leak and nerve root injury), and significant cost.7,9
In response, an ongoing prospective, multicenter, randomized controlled trial (RCT) is comparing a novel bone-anchored annular closure device (ACD, Figure 1) to conventional lumbar discectomy in patients classified as high-risk for reherniation (annular defect width ≥6 mm).10 That study reported significantly lower incidence of recurrent herniation and reoperation among ACD-treated patients during the first 90 days of follow-up. A previous direct cost analysis also suggested savings with the ACD11; however, a formal cost-utility analysis (CUA) has not been performed to determine cost-effectiveness.
Two approaches are commonly used to conduct a CUA: simple incremental calculation or decision analytical modeling.12 A major drawback of the cost-accounting approach is its inability to describe relationships between clinical events, impeding the prediction of how parameters change relative to one another. The purpose of this study was to perform a CUA via decision analytical modeling using a Markov method to evaluate the cost-effectiveness of the ACD versus conventional discectomies for patients with symptomatic lumbar disc herniation.
The patient population was extracted from a prospective, multicenter RCT.13 Included patients had single-level symptomatic lumbar disc herniations and underwent discectomy surgery. Randomization was performed intraoperatively following confirmation of an annular defect width ≥6 mm and patients received either limited discectomy or limited discectomy with insertion of the ACD. There were no statistically significant differences in baseline demographics.
Two primary outcome measures, cost and quality-adjusted life years (QALY) at 2 years follow-up, were evaluated using constructed health states. The resulting incremental cost-effectiveness ratio (ICER), expressed in dollars per QALY, was calculated and compared to willingness-to-pay (WTP) thresholds. The analysis was conducted in accordance with the Second Panel on Cost-Effectiveness Health and Medicine convened by the United States Public Health Service.14
Construction of Health States
All baseline and follow-up data were used to construct 5 health states that were based on 2 dimensions for pain and disability associated with symptomatic lumbar disc herniation: visual analog scale (VAS) and Oswestry Disability Index (ODI). These dimensions were combined using statistical regression to obtain a more comprehensive depiction of functional status. Multiple VAS measurements were obtained at each time point in the original trial (i.e., back pain, right or left leg pain), and were handled similarly to the original trial design. Health states were anchored by 20-point increments in maximum VAS. For ODI, health states are bound by the predicted values of ODI for each 20-point VAS increment (Figure 2).
Costs were measured from both healthcare (only direct medical costs) and societal (direct and indirect costs) perspectives. Humana and Medicare 2014 claims data were captured from a commercially available payer database (PearlDiver) to represent nationwide private and public payer data, respectively. Payer reimbursements were abstracted for pertinent diagnosis-related group (DRG) for inpatient scenarios, Ambulatory Payment Classifications (APC) for outpatient scenarios, and Current Procedural Terminology (CPT) codes for relevant procedures and services (Table 1).
For medication costs, we hypothesized that the type and number of medications taken immediately postoperatively should differ from later follow-up periods. In addition, this difference (particularly with analgesics, antispasmodics, and neuroleptics) was associated with the patient's health status (Table A1, http://links.lww.com/BRS/B364). We cross-referenced our data with the average wholesale price (AWP) from the 2011 Redbook MarketScan to calculate medication costs.15 Cost was estimated using AWP times 85%, based on Medicare's 2010 reimbursement rate for medications.
Costs associated with follow-up office visits (both scheduled and unscheduled) were also included in the model. Scheduled office visits occurred at 6 weeks, and at 3, 6, 12, 24, 36, 48, and 60 months, postoperatively. An analysis of the trial data revealed that unscheduled office visits were associated with a worse overall health status (P = 0.002). Therefore, we examined health state-specific likelihood of office visits per 6 weeks. All cost inputs were adjusted for inflation to 2017 dollars by using the US medical care Consumer Price Index.16
Productivity loss in this analysis was based on work status reported at each follow-up. In the Markov model, the states: “work with no restriction,” “work with few restrictions,” “work with many restrictions,” and “unable to work” were translated to 100%, 70%, 30%, and 0% work time, respectively. Productivity loss was not computed for patients older than 65. We applied 2016 US national average annual wages to calculate productivity loss (Appendix A2, http://links.lww.com/BRS/B364).17
The second primary outcome measure, health-related utility, was expressed in the unit QALY. The base case analysis utilized SF-36 data from the RCT and weighted utility values based on the validated SF6D algorithm.18 Both cost and QALYs were discounted at an yearly rate of 3% to reflect their present value.14 The cost-effectiveness outcome measure was calculated as the ICER. A value below the commonly accepted United States WTP threshold of $100,000 per QALY was considered to favor the ACD compared to conventional discectomy.14
A cohort Markov model (Figure 3) was constructed to analyze transition probabilities between health states for both ACD and control cohorts (Table A4, http://links.lww.com/BRS/B364). Each health state was therefore associated with costs and utility scores. The process of redistribution is controlled by 2 factors: the preoperative distribution of health states, and the transition probabilities between the health states. To capture the early dynamics of postoperative recovery, the model begins with 1.5-month cycles and then increases to 6-month cycles after the 6-month follow-up.
Four postoperative clinical events were considered in the model: reoperations related to symptomatic reherniation (repeat discectomy, decompression, and fusion); reoperation related to ACD failure (device revision and removal); reoperation unrelated to reherniation (decompression and fusion); and other types of complications (e.g., spinal cord stimulator placement, epidural hematoma evacuation, wound infection). Complications were associated with time-specific incidence rates and were included as input parameters to inform the model (Table 2).
The initial base case scenario assumed: a device cost of $3000; a 2-year time horizon; SF36 as the primary quality of life outcomes measure; direct costs were calculated using a 50:50 split between Medicare and Private payer rates; all index procedures occurred in the outpatient setting; and 60% of repeat discectomies were outpatient and 40% were inpatient.3
We conducted scenario, one-way parameter, and probabilistic sensitivity analyses. The scenario sensitivity analysis examined cost-effectiveness of the ACD versus “without ACD" in scenarios that deviated from the base case. For example, we tested alternate time horizons (i.e., 90 days, 1 year, and 5 years), reimbursement strategies (i.e., 100% Medicare or 100% private payer), and quality of life instruments (i.e., EQ5D). The purpose of the one-way parameter sensitivity analysis was to identify key parameters that influence the base case. We varied each input parameter, independently, by ±20%, and reported the resultant variations in the ICER in a tornado diagram (Figure 4). Parameters with the largest influence on the ICER are listed closest to the top of the illustration. The probabilistic sensitivity analysis examined the stability of the base case conclusion given the inherent uncertainty in the input parameters. We assigned beta distributions for probability and utility parameters and gamma distributions for cost parameters. Although uncertainty calculations are classically based directly on the 95% confidence intervals or standard errors, this level of granularity was not available for the cost data used in this analysis. Standard convention assumes that 95% of sampled costs fall into a range of ±20% or ±30% of their baseline values.19,20 To be conservative, we assumed a ±30% variation. We simulated 1000 iterations of input parameter variations (based on their respective distributions). The resultant cost and QALY differences between the ACD and control groups were illustrated on a scatter plot. Simulation points to the right of and below the blue US willingness-to-pay line indicates cost-effectiveness (Figure 5). A cost-effectiveness acceptability curve was also created from the probabilistic sensitivity analysis, which represents the percentage of iterations achieving cost-effectiveness over a range of willingness-to-pay thresholds (Figure 6).
A strong correlation (R 2 = 0.826, P < 0.001) was found by projecting VAS onto ODI for the creation of the Markov health states (Figure 2). Also, patients living in better health states were associated with a higher quality of life (Table 3) and increased likelihood of working with less restrictions (Table A2, http://links.lww.com/BRS/B364). Patients with the ACD had fewer symptomatic reherniations, reoperations, and complications and gained 0.0328 QALYs within the first 2 years.
The higher cost of index surgery in the ACD group was offset over time by reductions in complications, medications, and reoperations. At the conclusion of the base case time horizon (Table 4), total direct medical costs for the ACD group were essentially equal (+$198) to control. When productivity loss was considered, using the ACD became $2076 cheaper, per patient, than conventional discectomy.
From a health system perspective, the ICER comparing ACD to control equaled $6030 per QALY, a value that is substantially lower than the recommended US-based willingness-to-pay threshold of $100,000 per QALY. From a societal perspective, the ICER became negative—a rare scenario in healthcare economics referred to as dominance, which means that superior quality of life was attained at less cost.
Net monetary benefit was also analyzed and is a summary statistic that represents the value of an intervention in monetary terms when a willingness-to-pay threshold for a unit of benefit (i.e., QALY) is known. At 2 years and a willingness-to-pay threshold of $100,000/QALY, each ACD insertion saved $3084 of direct cost (Table 4). Moreover, when productivity loss was considered, the savings increased to $5358 per ACD insertion (Table 4).
From a healthcare perspective, with varying time-horizons (from 90 days to 5 years), the ICER comparing ACD to Control ranged from $2,272,750 per QALY to –$23,545 per QALY (Table 5). From a societal perspective, this range changed to $2,238,478 per QALY to a negative, dominant value. When costs were considered from an entirely Medicare fee schedule perspective, over the same follow-up time range, the ICER varied from $27,032 per QALY to dominating. And, when costs are considered from an exclusively private insurer's perspective, the ACD strategy dominated in all scenarios (Table 5). Using EQ5D instead of SF36 to assess quality of life did not significantly alter the result.
As the societal perspective strongly favored the ACD group, the one-way and probabilistic sensitivity analyses were confined to the health system perspective to be conservative. In the one-way sensitivity analysis, the parameters with the largest impact on the ICER included the reherniation rate in the control group, ACD device cost, and the cost of fusion surgery (Figure 4). Despite the resultant variations in the ICER, it consistently remained below $25,000 per QALY. The probabilistic sensitivity analysis also demonstrated that the majority (93.3%) of the iterations fall far below the standard willingness-to-pay threshold of $100,000 per QALY (Figures. 4 and 5). This remained valid (88.8% of iterations) even when a more antiquated willingness-to-pay threshold of $50,000 per QALY was considered.
Radicular low back pain from a herniated lumbar intervertebral disc remains one of the most common causes of disability and lost productivity in the United States.1–3 Despite the success of initial discectomy in most patients, reherniation-related reoperation costs are significant and place substantial stress on the healthcare system. It is estimated that $17 billion is spent annually on the readmission of medical and surgical patients in the USA.21 This constitutes approximately 20% of Medicare's total payments.22
In this study, we derived the ICER of ACD compared with traditional lumbar discectomy by using 2-year RCT data. For the base case analysis, an ICER of $6030 per QALY established cost-effectiveness centered on the commonly accepted willingness-to-pay threshold of $100,000 per QALY.14 Accounting for indirect costs augmented the ICER value further, resulting in economic dominance. A dominant healthcare economic strategy occurs when superior quality of life is attained at less cost.
Cost-effectiveness of the ACD compared to Control persisted despite varying input parameters in multiple sensitivity analyses, corroborating robustness of the model. In all scenarios, the ICER improved over time because of improving quality of life at less cost (Table 5). From a healthcare and Medicare-only perspective, the ICER increased to $27,032 per QALY. When costs were considered solely from a private insurer (Humana) setting, the ACD strategy dominated. In contrast, from a societal perspective, the ACD dominated conventional discectomy in both of these artificial monopolist payer scenarios. Similarly, in the probabilistic sensitivity analysis, cost-effectiveness could only be challenged if the simulated willingness-to-pay threshold was unrealistically low (i.e., at values <$10,000 per QALY).
The disparity between the healthcare and societal perspectives was believed to be secondary to differences in work status between the two groups. Health status-specific return-to-work probabilities suggested that patients with less postoperative disability were more likely to return to work (Table A2, http://links.lww.com/BRS/B364). Furthermore, return-to-work rates and health state transitions were superior in the ACD group (Tables A3 and A4, http://links.lww.com/BRS/B364). Before index surgery, only 13.1% of patients in the two groups could work without restriction. By 3, 12, and 24 months, the proportions of patients that could work without restriction in the ACD versus control groups were 47.8% versus 46.9%, 52.7% versus 49.2%, and 53.3% versus 49.5%, respectively (Table A3, http://links.lww.com/BRS/B364).
The ICER in this analysis compares favorably with that of other surgical interventions. A study of total ankle arthroplasty versus fusion demonstrated an ICER of $18,419 per QALY.23 An analysis of antibiotic-impregnated hip arthroplasty found an ICER of $37,595 per QALY24 and an examination of lumbar discectomy versus non-operative management calculated an ICER of $69,403 per QALY.25 Although these and other published reports have examined cost and quality of life in similar patient populations,26,27 this is the first study, to our knowledge, to create unique health states and use modeling to thoroughly assess the cost-effectiveness of an ACD. This novel approach allowed for extrapolation and accounted for the temporal component inherent in the disease process and patients’ recovery.
This study should be interpreted in the context of several limitations inherent in decision analytical approaches. By definition, the Markov model is conditional on the present state alone; future and past events are assumed independent. With disease processes, it is rarely plausible to assume that a patient's transition to another health state was entirely independent of a previous health state. The model also assumed that surgical cohorts began in similar health states, which is likely acceptable because of the trial randomization and nonsignificant differences in baseline characteristics.
We also recognize that some cost data were not ascertainable. As it is problematic to use hospital charge data to conduct a CUA, we used Medicare and Humana DRG/APC rates. As a result, differences in parameters (such as operating room time and length of stay) were not captured. Medication-related costs were estimated from the average wholesale price (0.85), as updated Medicare average sales prices were not publicly available. Although this estimate is considered appropriate, it is impossible to determine if it overestimated or underestimated costs for both groups. Productivity loss was also a significant contributor to cost, but this analysis was unable to include factors such as transportation costs, caregiver time/responsibilities, and educational days missed.
This study is the first to report the comparative cost-effectiveness of a novel ACD compared to conventional lumbar discectomy using a decision analytic model. The authors conclude that for patients with lumbar disc herniations, the ACD used in the trial is, at 2 years, a highly cost-effective surgical modality compared to conventional lumbar discectomy. Furthermore, cost savings and quality of life both continued to improve over time. In a rapidly changing medical climate with emerging practice paradigms such as pay-for-performance and value-based purchasing, surgeons and payers will naturally gravitate toward these analyses. In addition, as the health care system becomes more informed in an established setting of scarce resources and increasing expenses, sustainable surgical technologies that improve quality of life while saving costs require serious attention.
1. Frymoyer JW. Back pain and sciatica. N Engl J Med
2. Vroomen PC, de Krom MC, Knottnerus JA. Predicting the outcome of sciatica at short-term follow-up. Br J Gen Pract
3. Vroomen PC, de Krom MC, Wilmink JT, et al. Lack of effectiveness of bed rest for sciatica. N Engl J Med
4. Peul WC, van Houwelingen HC, van den Hout WB, et al. Surgery versus prolonged conservative treatment for sciatica. N Engl J Med
5. Gray DT, Deyo RA, Kreuter W, et al. Population-based trends in volumes and rates of ambulatory lumbar
spine surgery. Spine (Phila Pa 1976)
6. Carragee EJ, Spinnickie AO, Alamin TF, et al. A prospective controlled study of limited versus subtotal posterior discectomy: short-term outcomes in patients with herniated lumbar
intervertebral discs and large posterior anular defect. Spine (Phila Pa 1976)
7. Ambrossi GL, McGirt MJ, Sciubba DM, et al. Recurrent lumbar disc herniation
after single-level lumbar
discectomy: incidence and health care cost analysis. Neurosurgery
8. McGirt MJ, Eustacchio S, Varga P, et al. A prospective cohort study of close interval computed tomography and magnetic resonance imaging after primary lumbar
discectomy: factors associated with recurrent disc herniation
and disc height loss. Spine (Phila Pa 1976)
9. Parker SL, Mendenhall SK, Godil SS, et al. Incidence of low back pain after lumbar
discectomy for herniated disc and its effect on patient-reported outcomes. Clin Orthop Relat Res
10. Klassen PD, Bernstein DT, Köhler HP, et al. Bone-anchored annular closure following lumbar
discectomy reduces risk of complications and reoperations within 90 days of discharge. J Pain Res
11. Parker SL, Grahovac G, Vukas D, et al. Cost savings associated with prevention of recurrent lumbar disc herniation
with a novel annular closure device
: a multicenter prospective cohort study. J Neurol Surg A Cent Eur Neurosurg
12. Ament JD, Kim KD. Standardizing cost-utility analysis
in neurosurgery. Neurosurg Focus
14. Sanders GD, Neurmann PJ, Basu A, et al. Recommendations for conduct, methodological practices, and reporting of cost-effectiveness
analyses. Second panel on cost-effectiveness
in health and medicine. JAMA
15. Marketscan. Commercial Claims and Encounters and Medicare Supplemental Databases. Available at: http://redbook.com/redbook/awp//
. Accessed September 6, 2017.
16. Medical Consumer Price Index. Bureau of Labor Statistics. Washington, DC: United States Department of Labor; 2016. Available at: http://www.bls.gov/cpi/cpifact4.htm
. Accessed September 1, 2017.
19. Picot J, Cooper K, Bryant J, et al. The clinical effectiveness and cost-effectiveness
of bortezomib and thalidomide in combination regimens with an alkylating agent and a corticosteroid for the first-line treatment of multiple myeloma: a systematic review and economic evaluation. Southampton (UK): NIHR Journals Library; 2011 Dec. (Health Technology Assessment, No. 15.41.) Appendix 14, Parameters included in the sensitivity analyses. Available from: https://www.ncbi.nlm.nih.gov/books/NBK97492/
20. Loveman E, Jones J, Clegg AJ, et al. The clinical effectiveness and cost-effectiveness
of ablative therapies in the management of liver metastases: systematic review and economic evaluation. Southampton (UK): NIHR Journals Library; 2014 Jan. (Health Technology Assessment, No. 18.7.) Appendix 11, Variables included in probabilistic sensitivity analysis, sampling distributions and parameterisation of sampling distributions. Available from: https://www.ncbi.nlm.nih.gov/books/NBK261554/
21. Jencks SF, Williams MV, Coleman EA. Rehospitalizations among patients in the Medicare fee for-service program. N Engl J Med
22. Medicare Payment Advisory Commission (2007) Report to the Congress: promoting greater efficiency in Medicare. Medicare Payment Advisory Commission.(MedPAC).
23. SooHoo NF, Kominski G. Cost-effectiveness
analysis of total ankle arthroplasty. J Bone Joint Surg Am
24. Cummins JS, Tomek IM, Kantor SR, et al. Cost-effectiveness
of antibiotic-impregnated bone cement used in primary total hip arthroplasty. J Bone Joint Surg Am
25. Tosteson ANA, Skinner JS, Tosteson TD, et al. The cost effectiveness of surgical versus nonoperative treatment for lumbar disc herniation
over two years: evidence from the Spine Patient Outcomes Research Trial (SPORT). Spine (Phila Pa 1976)
26. Walid MS, Robinson JS Jr. Economic impact of comorbidities in spine surgery. J Neurosurg Spine
27. Auffinger BM, Lall RR, Dahdaleh NS, et al. Measuring surgical outcomes in cervical spondylotic myelopathy patients undergoing. PLoS One
2013; 8: e67408.