This study was conducted in compliance with the principles of the Declaration of Helsinki. The study protocol was reviewed and approved by the Siriraj Institutional Review Board (IRB), Faculty of Medicine Siriraj Hospital, Mahidol University (IRB No. Si 700/2021). As this research did not include patient-specific data, the requirement for informed consent was waived in accordance with local regulations.

Patients with SLS were enrolled prospectively and retrospectively between 2014 and 2022. Eligible participants were adults aged ≥18 years with single-level degenerative LSS confirmed by radiographic studies and surgical indications, including progressive neurological deficits or severe radicular pain despite ≥3 months of conservative treatment. In this cohort, however, all enrolled patients were middle-aged or elderly, with median ages of 60.4 years in the PLF/TLIF group and 63.9 years in the PLF group (Table 1), and an age range of 41–83 years. Although an inclusion criterion allowed adults aged ≥18 years to reflect real-world surgical practice, very young adults were rarely represented in our sample. Consequently, the study population predominantly comprised typical degenerative LSS cases in older adults. Instrumentation was performed for mechanical instability or severe canal stenosis, ensuring inclusion of degenerative pathology. Radiologically confirmed instability or ventrolisthesis was included if a surgical indication was present. Patients with multilevel or tandem stenosis were excluded, and sagittal alignment was not evaluated. Additional exclusion criteria included a history of previous lumbar surgery and the presence of mental illness, dementia, or altered consciousness. A retrospective review of imaging confirmed that the anatomical distribution of spinal stenosis, including central, lateral recess, and foraminal, was similar across groups. All eligible patients who consented to participate were consecutively enrolled. Treatment was guided by clinical judgment and shared decision-making. All surgeries were performed by board-certified spine surgeons with more than 10 years of independent surgical experience. The impairment scores for single spinal canal stenosis, as reported by Carreon et al. [5], were used for sample size calculation. The mean disability index, measured by the ODI, was 30.4±18.6 for 101 patients who underwent PLF/TLIF, compared with 21.1±19.2 for the PLF group. A statistical formula for testing two independent means was employed, and the study required a minimum sample size of 63 patients per group.

Patients were divided into two groups: PLF/TLIF and PLF. In the PLF/TLIF group, surgery included bilateral pedicle screw placement, followed by unilateral TLIF with a PEEK (polyether ether ketone) cage packed with local autograft bone obtained from the decompression (lamina, spinous processes, and facet joints). Bilateral rod fixation and final posterolateral fusion using the same local autograft were then completed. In the PLF group, a similar approach with bilateral pedicle screw and rod fixation and posterolateral fusion using only local autograft bone was performed, without additional iliac crest bone harvesting. Perioperative allogeneic blood transfusion was determined according to the attending anesthesiologist’s and surgeon’s clinical judgment, considering estimated blood loss, hemoglobin level, hemodynamics, and comorbidities. Preoperative, perioperative, and 1-year postoperative data were collected. Perioperative outcomes included estimated blood loss, operative time, and hospital stay. Quality of life and functional outcomes were evaluated using the EuroQol 5-dimension 5-level (EQ-5D-5L), EuroQol-Visual Analog Scale (EQ-VAS), and ODI. The dimensions of EuroQol-5 Responses across the five levels (EQ-5D-5L) were translated into utility scores using coefficient factors tailored to the Thai population [7]. This combined construct (unilateral TLIF with an interbody cage plus posterolateral fusion) is referred to as PLF/TLIF throughout this study.

Clinical and demographic characteristics were described and evaluated. Continuous and categorical data were analyzed using an independent t-test and a chi-square test, respectively. The mean and standard deviation of the results were presented. Statistical significance was defined as a two-tailed p-value of less than 0.05. Several data analyses were conducted using IBM SPSS Statistics for Windows ver. 19.0 (IBM Corp., Armonk, NY, USA), and Microsoft Excel 2019 (Microsoft Corp., Redmond, WA, USA) was used for economic evaluations.

The cost-effectiveness analysis (CEA) was based on 1-year outcome scores from the ODI. CEA was presented as the incremental cost-effectiveness ratio (ICER), comparing PLF/TLIF and PLF by analyzing the ratio of increased societal costs to increased effectiveness. In this study, the ICER was calculated as (C₁−C₂)/(E₁−E₂), using the ODI as the measure of effectiveness.

The cost-utility analysis compared the lifetime costs and health outcomes of the PLF/TLIF group with those of the PLF group. The economic evaluation was conducted following the 2022 Consolidated Health Economic Evaluation Reporting Standards guidelines [6]. The natural course of the disease in patients was modeled using a Markov model and a decision tree. According to the Thai Health Technology Assessment criteria, the cost-utility results were evaluated from a societal perspective using a lifetime horizon [10].

The findings are shown as net monetary benefit (NMB) and ICERs, expressed in the Thai baht (THB) per quality-adjusted life-year (QALY) gained. Thailand’s willingness-to-pay (WTP) threshold of US$5,003 per QALY gained, with a 3% discount rate, was used for the cost-utility analysis. The average 2022 exchange rate of US$1=31.98 THB was applied for all cost conversions.

The economic model was initiated with a decision tree, as illustrated in Fig. 1A. Patients were then transitioned into a five-health-state Markov model capturing lifetime costs and health outcomes. The five health states encompassed “well,” “complications,” “index revision,” “adjacent revision,” and “death” (Fig. 1B). Arrows represent patient flow, showing how patients remained in their current state or transitioned to another state.

Data on the course of the disease and surgical methods described by Pannell et al. [11], Levin et al. [12], Carreon et al. [5], and Adogwa et al. [13] were used to calculate transition probabilities. From a societal perspective, costs included direct nonmedical costs (travel and meals) and direct medical costs (lodging, medication, nursing, imaging, surgery, anesthesia, implants, and physical therapy). The QALY analysis was expected to show disutility for any loss of work capacity or leisure activities resulting from disease or treatment [14]. Consequently, indirect expenses were excluded from the analysis. All model input parameters are presented in Table 2. The revision rates from published studies were used as input parameters in the Markov model. These values do not reflect cohort revisions but represent long-term transition probabilities.

A retrospective data review was used to collect utility statistics, outpatient and hospital visit rates, and direct treatment expenses. In compliance with the Thai health technology evaluation criteria, direct nonmedical expenditures were taken from Thailand’s Standard Cost Lists for Health Economic Evaluation [15]. Prospectively, utility data were gathered at the preoperative, perioperative, and 1-year postoperative intervals.

In the base-case analysis, QALYs and total lifetime costs of PLF/TLIF were compared with those of PLF alone. One-way sensitivity analyses were conducted to evaluate the impact of modifying input parameters within 95% confidence intervals (CI) ranges. In the one-way sensitivity analysis, clinical effects, transition probabilities, costs, and utilities were varied within their 95% CI ranges. Probabilistic sensitivity analysis (PSA) examined combined parameter uncertainty. Costs were assigned gamma distributions, and utilities and transitional probabilities were assigned beta distributions. To estimate total lifetime costs and QALYs, 1,000 Monte Carlo simulation iterations were performed. The PSA results were illustrated using a cost-effectiveness plane.

Additional multivariable regression analyses were conducted to account for potential confounding by age, systemic comorbidities, and surgical level. Separate linear regression models were fitted with 12-month ODI and 12-month EQ-5D-5L index scores as dependent variables, including type of surgery (PLF/TLIF vs. PLF), age, diabetes mellitus, hypertension, dyslipidemia, rheumatoid arthritis, chronic kidney disease, other comorbidities, and L5–S1 involvement as covariates. A log-linear model was used to evaluate the association between these covariates and short-term direct medical costs.

This study included 140 patients diagnosed with SLS who underwent surgery, with 70 patients in each group. The mean age was 60±11.7 years in the PLF/TLIF group and 64±10.5 years in the PLF group. Patient demographics and clinical characteristics are summarized in Table 1. Baseline characteristics, including sex, age, body mass index, underlying disorders, diagnoses, levels of operation, American Society of Anesthesiologists scores, and length of hospital stay, did not differ significantly between the two groups (p>0.05). L4–5 was the most commonly involved level in both groups, whereas L5–S1 involvement was more frequent in the PLF/TLIF group than in the PLF group (15/70 [21.4%] vs. 4/70 [5.7%]; Fisher’s exact test, p=0.012) (Table 1). Estimated blood loss did not differ significantly between the PLF and PLF/TLIF groups (267.21 mL vs. 281.57 mL; p=0.49). However, the operative time was significantly longer in the PLF/TLIF group (160.96 minutes vs. 138.86 minutes; p=0.002). At the 1-year follow-up, both groups demonstrated improvements in ODI and EQ-VAS scores. The mean 12-month ODI was lower in the PLF/TLIF group than in the PLF group (22.24±15.85 vs. 37.65±57.69), with a mean difference of −15.41 points (95% CI, −29.4 to −1.4; p=0.018), indicating superior functional improvement with PLF/TLIF. In contrast, 12-month EQ-5D-5L utility scores were similar between the groups (0.83±0.23 vs. 0.84±0.17; mean difference, −0.01 [95% CI, −0.08 to 0.06]; p=0.466). The data are shown in Table 3. No cases of revision surgery, hardware loosening, or mechanical failure were observed in either group during the 12-month follow-up. Overall, these findings indicate that PLF/TLIF provided superior functional improvement compared with PLF, despite requiring a longer operative time and comparable perioperative blood loss.

In multivariable models adjusting for age, diabetes mellitus, hypertension, dyslipidemia, rheumatoid arthritis, chronic kidney disease, other comorbidities, and L5–S1 involvement, type of surgery remained associated with lower 12-month ODI in favor of PLF/TLIF, while 12-month EQ-5D-5L scores remained similar between groups. In the cost model, PLF/TLIF remained associated with higher short-term direct medical costs, approximately 14% greater than PLF. Diabetes mellitus demonstrated a modest positive association with costs, whereas L5–S1 involvement and other comorbidities were not significant predictors. Adjustment for these covariates did not materially alter the incremental differences in disability, quality of life, or costs between PLF/TLIF and PLF.