How Decompression Surgery Improves the Lower Back Pain in Patient with Lumbar Degenerative Stenosis

A Propensity-Score-Matched Analysis

Mitsuru Yagi, MD, PhD; Satoshi Suzuki, MD, PhD; Satoshi Nori, MD, PhD; Yohei Takahashi, MD, PhD; Osahiko Tsuji, MD, PhD; Narihito Nagoshi, MD, PhD; Masaya Nakamura, MD, PhD; Morio Matsumoto, MD, PhD; Kota Watanabe, MD, PhD


Spine. 2022;47(7):557-564. 

In This Article

Materials and Methods


No external funding was used for this study. The authors report no conflicts of interest.

Patient Enrolment

We conducted a retrospective analysis of data entered prospectively into a single-center database. This study was approved by the institutional review board of our hospital (IRB approval number #20110142). We used a prospective database from a single academic hospital to analyze data from 436 consecutive patients who were surgically treated for LSS by posterior decompression alone (laminotomy) between April 2012 and March 2018 and who had 2 years of follow-up data. All data were collected prospectively and analyzed retrospectively.

Inclusion and Exclusion Criteria

This study included adult patients (age ≥ 21 yrs) diagnosed with primary LSS, with or without lumbar degenerative spondylolisthesis (DS, Meyerding grade 1), and with clinical symptoms and ineffective conservative treatment for at least 3 months or recurrent symptoms. We chose decompression surgery only when the patient did not have spinal instability on flexion-extension dynamic radiographs before surgery. Patients were excluded if they were followed up for less than 2 years after surgery, had a previous spinal surgery, or had an incomplete dataset.

Data Preparation

As a surrogate for LBP intensity evaluation, the presence of persistent postoperative substantial LBP was defined as a visual analogue scale (VAS) score ≧4.5 at the 2-year postoperative follow-up evaluation according to the previously described threshold, while severe LBP was defined as a VAS score ≧7.5.[20] For detailed analyses, we compared the proportions of patients achieving a minimum clinically important difference (MCID) for LBP. The MCID for LBP on the VAS score was set to 2.0 based on the previous literature.[21] Similarly, the minimal symptom state (MSS) was defined as a VAS score for LBP less than 2.0.[22]

Relationships between 2-year postoperative substantial LBP and the baseline patient background factors, surgical factors, and health-related quality of life (HRQOL) were investigated by univariate and multivariate logistic regression analyses, as well as propensity score-matched analyses.

Data Collection and Radiographic Data and HRQoL Data Assessment

The collected demographic and clinical data included patient age at index surgery, sex, diagnosis, comorbidities, frailty, smoking, job type, number of intervertebral levels involved, time of surgery, amount of estimated blood loss, perioperative complications, and incidence of revision surgery.

The two types of patient-reported outcomes (PROs), including the visual analogue scale (VAS) scores for lower back pain, buttock-leg pain, and leg numbness and Japanese Orthopaedic Association Back Pain Evaluation Questionnaire (JOABPEQ) scores, were collected at baseline and at the 1- and 2-year postoperative follow-up evaluations to assess changes in individual HRQoL. The JOABPEQ is a PRO instrument developed by the Japanese Orthopaedic Association (JOA) for the assessment of lower back pain and lumbar spinal disease, and the JOABPEQ has been validated in a number of countries, including the United States, Korea, Thailand, China, and Iran.[23–30] In the present study, among the 436 patients, all of the patients had baseline and 2-year postoperative follow-up evaluation VAS and JOABPEQ scores. No disease-related deaths occurred within 2 years after surgery.

Unadjusted and Adjusted Risk Analysis for Substantial Lower Back Pain at the 2-year Postoperative Follow-up Evaluation and Baseline Severe Lower Back Pain in the Patient Cohort

In the present study, we analyzed the risk of substantial LBP at the 2-year postoperative follow-up evaluation and baseline severe LBP. After the descriptive analysis, independent associations between potential risk factors and lower back pain were analyzed by univariate comparison. Univariate risk analyses were carried out with unpaired t tests and Tukey honest significant difference test or the Wilcoxon ranked test where appropriate.[31] To compare the odds ratio (OR) of substantial postoperative lower back pain, we created a multivariate logistic regression model to evaluate the adjusted associations of each potential explanatory variable and to predict the likelihood of lower back pain.

Propensity Score Matching of the Patient Cohort

To account for the potential confounding factors for sex differences, patients from the databases were propensity score matched for age, LBP aetiology, frailty, levels treated, baseline JOABPEQ mental health domain score, and the severity of baseline lower back pain (VAS score).[31] There were 204 patients (102 patients in each group) who were matched by propensity scoring (Table 1). The chi-square of the Hosmer-Lemeshow test for this propensity score matching was 7.849, and the P value was 0.448, indicating good model adaptation.

Statistical Analysis

We calculated the overall summary statistics, including the means and standard deviations for continuous variables and the frequencies and percentages for categorical variables. Changes between baseline and postoperative values were analyzed by paired t tests. Differences between 2-year postoperative lower back pain in patients with VAS scores < 4.5 and ≥4.5 and differences in baseline lower back pain in patients with VAS scores <7.5 and ≥7.5 were compared by the chi-square test, an analysis of variance (ANOVA), or Tukey-HSD test where appropriate. The mean difference between the above-mentioned groups was calculated with a 95% confidence interval (CI). A P value less than 0.05 with a CI of 95% was considered statistically significant. After the chi-square test, residual analyses were carried out to identify those specific cells making the greatest contribution to the chi-square test result.[31]

The outcomes of interest included (1) age, (2) sex, (3) job type, (4) LBP aetiology, (5) smoking history, (6) functional status, (7) modified frailty index (mFI), (8) the type and number of comorbidities, (9) a history of joint replacement, (10) estimated blood loss, (11) surgical time, (12), levels decompressed, (13) the development of perioperative complications, (14) the presence of spinal instability at the 2-year postoperative evaluation (15) the presence of spondylolisthesis at the 2-year postoperative evaluation, and (16) baseline JOABPEQ mental health domain score. Categories were created as follows: age every 10 years and frailty robust (mFI = 0), prefrail (mFI < 0.21), or frail (mFI ≥ 0.21). A P value less than 0.05 was considered statistically significant. Data were analyzed with Statistical Package for the Social Sciences software (SPSS statistics version 27.0, IBM Corp, Armonk, NY).