Patients and setting

This economic evaluation was conducted alongside the PTED-study, a large multicentre randomised controlled trial with a non-inferiority design.

A detailed description of the PTED-study, including its sample size calculation, has previously been published.20 In brief, participants were recruited between February 2016 and April 2019 from three hospitals and one private health clinic in the Netherlands. To be eligible, patients had to meet the following inclusion criteria: age between 18 and 70 years; >10 weeks of radiating pain with or without motor or sensory loss in the leg or with >6 weeks of excessive radiating pain and no tendency for any clinical improvement; indication for surgery; MRI-confirmed lumbar disc herniation with nerve compression with or without concomitant spinal or lateral recess stenosis; sufficient knowledge of the Dutch language. Exclusion criteria were previous surgery on the same or adjacent disc level; cauda equina syndrome; spondylitis or degenerative spondylolisthesis; pregnancy; severe comorbid medical or psychiatric disorder (American Society of Anesthesiologists >2); severe caudal or cranial sequestration; contraindication for surgery; moving abroad at short notice. Patients with lumbar disc herniation were screened for eligibility during their outpatient consultation by one of the participating surgeons. Eligible patients received written information about the study and were given at least 2 days to consider participation. After that period, a trained research nurse further screened patients who were willing to participate, informed consent was obtained, and baseline measurements were performed. Then, patients were randomised in a 1:1 ratio to PTED or OM, using a computer-generated, random-number tables with variable block sizes (ie, 4, 6, 8), stratified by treatment centre. Treatment allocation was concealed and was performed by an independent research nurse. Blinding of patients was not possible due to the fundamentally different nature of both procedures. Outcome assessors were not blinded either, because all outcomes were self-reported.20

Interventions

The PTED procedure has been extensively described elsewhere.11 In brief, PTED was performed under local anaesthesia and conscious sedation, on an outpatient basis. By means of fluoroscopy and a guidewire, conical rods were introduced into the neuroforamen. Subsequently, drills were introduced to enlarge the neuroforamen, followed by an endoscope and rongeur to remove the disc fragments.

Open microdiscectomy was performed under general anaesthesia. The use of a microscope or surgical loupes was optional. A paramedian incision was made. Following the identification of the lamina, the yellow ligament was removed to identify the nerve root and disc herniation. Laminotomy, as well as foraminotomy, was performed, if necessary. For the foraminal herniated disc, a partial medial facetectomy was used, and for the extraforaminal herniated disc, an approach alongside the facet joints was used. The patient was discharged as soon as medically responsible, which is usually 1 day after surgery.

Effect measures

The primary effect measures for the economic evaluation were intensity of leg pain measured on the Visual Analogue Scale (VAS) ranging from 0 (no pain) to 100 (worst imaginable pain) and quality-adjusted life years (QALYs), derived from the EQ-5D-5L.20 QALYs, a commonly used outcome in economic evaluation, are a generic measure which provides a common outcome metric across clinical areas enabling comparisons across conditions. We also included pain intensity because it was the primary clinical outcome of the effectiveness trial and facilitates comparison of the results within the Spine-field. Both measures were assessed at baseline, the day after surgery, at 2, 4 and 6 weeks, and 3, 6, 9, 12 months. The EQ-5D-5L measures health-related quality of life in five health dimensions: mobility, self-care, daily activities, pain/ discomfort and anxiety/depression. The Dutch EQ-5D-5L tariff was used to convert the patients’ EQ-5D-5L health states into utility scores, anchored by the health states of death (0) and perfect health (1.00).22 QALYs were estimated using the ‘area under the curve’, meaning that QALYs were calculated by multiplying the patients’ health state utility scores with the time spent in that health state. A linear relationship between the patients health state utility scores at the various time points was assumed.23 The total QALY of the 1-year follow-up period was calculated by adding up the QALYs for each follow-up period (0–2 weeks, 2–4 weeks, 4–6 weeks, etc) assuming a linear increase in QALY within each period.

Cost measures

Resource use was assessed using cost questionnaires administered at baseline and at 2, 4 and 6 weeks, and 3, 6, 9, 12 months. Since this study adopted a social perspective, both direct and indirect costs were included. Direct costs included costs of the intervention, primary healthcare use, secondary healthcare use and the use of medication. Intervention costs were estimated using hospital accounting records, while all other healthcare utilisation was valued according to the Dutch guidelines24 and medication use using prices of the Dutch Health Care Institute (http://medicijnkosten.nl). Costs of the interventions include the time of the operating room used, the costs of the medications used during the surgery and for open microdiscectomy also the cost for one overnight hospital stay. Indirect costs consisted of absenteeism, presenteeism, unpaid productivity and informal care costs. Absenteeism was assessed by asking patients to report their number of sick days and valuing them using the friction cost method (friction period=12 weeks) with wages adjusted for gender.24 Presenteeism (ie, reduced productivity at work) was assessed using the World Health Organization-Health and Work Performance Questionnaire and valued using the same gender-specific wages. Unpaid productivity losses (ie, losses due to being unable to perform unpaid activities, such as volunteer work) and informal care (ie, care by family and friends) were valued using a recommended Dutch shadow price (€15). All costs were converted to euros 2019 using consumer price indices. Discounting of costs was not necessary.

Confounding variables

Based on the literature,25 clinical experience and consensus among the project team, the following potential confounders were identified: age (years), gender (male/female), smoking status (yes/no), body mass index (weight/height,2 employment status (yes/no), duration of complaints (months), morphological location of disc herniation (intraforaminal/extraforaminal), psychopathology as measured with the Four-Dimensional Symptom Questionnaire,26 and treatment preferences (no/PTED/open microdiscectomy).

Statistical analyses

A CE analysis (CEA) and a cost-utility analysis (CUA) were conducted. In the CEA, total costs were related to improvement of leg pain during 12-month follow-up. In the CUA, total costs were related to QALYs gained during follow-up. The primary analysis was conducted according to the intention-to-treat approach. All missing data were imputed using multivariate imputation by chained equations.27 The imputation was stratified by treatment group to account for association of the treatment group with missingness.28 To deal with the association between being part of the learning curve group (yes/no) and the missingness of data, data from learning curve patients were excluded before imputing data for the main analysis and five sensitivity analyses. Predictive mean matching was used to create ten complete datasets. Disaggregate cost differences were analysed using linear regression models, both adjusted and unadjusted for confounders. Differences in total costs and effects between treatment groups were obtained from a system of seemingly unrelated regressions that accounted for the potential correlation between costs and effects.29 These total cost and effect differences were adjusted for baseline and confounders. Incremental cost-effectiveness ratios (ICERs) were calculated by dividing incremental costs by incremental effects. Uncertainty surrounding ICERs and cost differences were estimated by bias corrected and accelerated bootstrapping with 5000 repetitions. Uncertainty was presented using CE Planes and cost-effectiveness acceptability curves (CEACs).30 31 Results were pooled over the imputed datasets using Rubin’s rules.27 Analyses were performed in R statistical programming language, V.3.6.1.

Sensitivity analyses

To assess the robustness of the results, six sensitivity analyses were performed for both effect measures. First, a crude analysis was run (unadjusted for potential confounders). Second, only complete cases were used. Third, two scenarios of PTED intervention costs were considered. In the high-cost scenario, the cost of PTED was €5000/patient, that is, €500 more than in the main analysis. In the low-cost scenario, the cost of PTED and open microdiscectomy were equal, that is, €4095/patient. Fourth, productivity losses were measured according to the human capital approach. Fifth, the healthcare perspective was adopted. Sixth, learning curve patients were included. Except for the high-cost and low-cost scenari’s, all sensitivity analyses were planned a priori.

Patient involvement

Prior to the start of the PTED-study, the proposed study design and study procedures were presented to the relevant physician organisations, members of the Dutch Health Insurance board and members of the patient organisation ‘de Wervelkolom’ (translated: the Spine). Based on the input of these organisations on aspects such as feasibility, patient friendliness and patient safety, the study design would be updated if necessary. Furthermore, the organisations such as the patient organisation were invited to be part of the half-yearly board meetings during which recruitment, implementation and results of the study were discussed.