2025 Volume 7 Issue 1 Pages 1-9
BACKGROUND
Reoperation after lumbar spine surgery is a major issue for both patients and physicians. It is uncertain whether fusion is superior to decompression alone for lumbar degenerative disease regarding reoperation rate. We aim to evaluate the reoperation rate after fusion surgery for lumbar degenerative disease compared with decompression alone.
METHODS
This study was conducted under a retrospective cohort design in patients undergoing fusion or decompression alone in one or two levels for lumbar degenerative disease using a Japanese claims-based database. Primary outcome was reoperation incidence during the follow-up period, and secondary outcome was reoperation incidence within 90 days postoperatively. Confounding factors were handled using propensity score overlap weighting. Cumulative incidence of reoperation was calculated from the Kaplan-Meier curve and hazard ratios (HRs) and 95% confidence intervals (CIs) for reoperation were estimated using Cox proportional hazards regression models.
RESULTS
8497 patients (2051 patients in the fusion group and 6446 in the decompression alone group) were included in the study. There was no difference in reoperation rate between fusion and decompression alone (weighted HR 0.85 [95% CI 0.69 to 1.04]; p = 0.11).
CONCLUSIONS
Among patients with lumbar degenerative disease who underwent fusion or decompression alone, no significant difference was observed between the two groups.
Surgical treatments for lumbar degenerative disease (LDD) are broadly classified into decompression alone and fusion. Although the efficacy of decompression alone has been shown, it does not treat the instability of LDD, which could result in postoperative pain1,2). During the past few decades, fusion has been increasingly added to decompression for LDD to treat the instability at the affected level3–6). While fusion is expected to improve symptoms arising from LDD and decrease the potential risk of future instability and deformity, potential disadvantages include increased invasiveness, higher cost, and adjacent segment degeneration7,8). The question of which surgical method is better treatment for LDD remains controversial and practice varies widely3,6,9).
The reoperation rate after LDD surgeries varies from 5% to 19% on 3 to 15 years’ follow-up10–13). Reoperation is a detrimental event, which leads to patient dissatisfaction14). Furthermore, reoperation is associated with a worsening of clinical outcomes, such as pain, neurological deficit, and quality of life15).
Among a number of studies that have compared the reoperation rates of fusion and decompression alone, all but one—a randomized controlled study—reported no difference in reoperation rate between them11–13,16–19). However, all these studies were limited in sample size, insufficient length of follow-up, exclusion of reoperation in other facilities, or investigated the reoperation rates of LDD surgeries performed in the early 1990s to 2000s. As surgical methods and trends for LDD have dramatically changed3,4,6,9), the need to reevaluate the reoperation rate of fusion compared with decompression alone is compelling.
In this study, we analyzed data from a large claims-based database to investigate the incidence of reoperation following fusion compared with decompression alone. We hypothesized that in real-world daily clinical practice, fusion results in a decreased risk of reoperation compared with decompression alone.
The study was approved by the ethical committee of our institution (No. R3551). It followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline20). Individual informed consent is waived because all patient information is anonymized.
STUDY DESIGN AND SETTINGThis retrospective cohort study was conducted using a commercially available administrative claims database from JMDC Inc.21). This database included approximately 14 million patients as of 2021 and is a major source of medical claims data22,23). It is sourced from health insurance societies for company employees and their dependents in Japan. It is limited to patients aged 74 or younger because it does not cover claims associated with the late-stage medical care system for those aged 75 or older. The database includes patient characteristics, diagnoses, procedures, drugs, materials, and specific health checkup records21,24). Individuals can be followed through a unique encrypted identifier even if they visit or are hospitalized at multiple medical institution, enabling precise evaluation of reoperation rates23).
PARTICIPANTSOur study population consisted of patients who underwent fusion or decompression alone for LDD between Jan 2010 and Oct 2019 and were followed up to Oct 2021. The date of fusion or decompression alone was set as the index date. Exclusion criteria were patient age <18 years at the index date, surgery in three or more levels, combined fusion and decompression alone at the other intervertebral levels, a medical history of less than one year before and 90 days after the index date in this database, history of lumbar surgery in the year preceding the index date, and concomitant diseases (lumbar spinal fracture, spinal infection, or spinal tumor). We excluded patients those who underwent fusion or decompression in three or more levels, as including these patients may lead to significant selection bias. Detailed definitions of inclusion and exclusion criteria are shown in the Supplementary Table 1. We set a minimum follow-up duration of 90 days because excluding patients with less than two years’ follow-up may limit our study population who can continue to be enrolled in the same insurance system, which may bias the study population towards healthier people.
EXPOSUREThe main exposure variable was fusion versus decompression alone, whichever occurred earlier. Fusion included posterior fusion, posterior interbody fusion, anterior interbody fusion, and anterior-posterior fusion12,19). Decompression alone included any procedure involving laminectomy or discectomy without arthrodesis (Supplementary Table 1)12,19). If some patients experienced both fusion and decompression alone during the follow-up, we regarded the earlier surgery as the primary surgery and the later one as a reoperation. Both the sensitivity and specificity for these administrative data compared with medical record review were over 95% for both fusion and decompression alone25).
OUTCOMESThe primary outcome was the incidence of any type of second surgery including debridement for surgical site infection and postoperative epidural hematoma during the follow-up period (Supplementary Table 2)11,19). This is because the reasons for reoperation after fusion and decompression alone differ: reoperation after fusion is performed because of the failure to achieve osseous fusion, adjacent segment degeneration, or persistent pain, whereas reoperation after decompression alone is conducted due to insufficient decompression, worsening instability, or re-stenosis10,16). Therefore, including reoperations at every lumbar spine level allows us to evaluate the reoperation risk of fusion compared with decompression alone. Patients were followed from the index date and until the first occurrence of a study outcome, death, disenrollment, end of the study period, whichever came first. The secondary outcome was the incidence of any lumbar spine surgery within 90 days postoperatively. This is because the reason for reoperation before and after 90 days may differ19): while reoperation before 90 postoperative days may be caused by insufficient decompression, wrong level of operation, or implant failure, reoperation after 90 days postoperatively may be related to the progression of degenerative disease, instability, or nonunion.
COVARIATESThe following data were extracted from the database: age, gender, body mass index (BMI), smoking status, comorbidities (diabetes mellitus, hemodialysis, rheumatoid arthritis, osteoporosis, any malignancy, and Charlson Comorbidity Index [CCI])26), type of hospital, and surgery year. We split the surgery years into 2010–2014 and 2015–2019 to account for the surgical trends and the learning curves. These covariates were determined based on our clinical experience and previous reports27–30). Detailed definitions of covariates are shown in the Supplementary Table 3.
STATISTICAL ANALYSISPropensity score (PS) overlap weighting analysis was used to adjust for the measured confounders listed above between the two groups31). We could not include BMI and smoking status because these data were missing for a substantial number of patients. A PS for receiving fusion (i.e., the probability that a patient would receive fusion) was created using a logistic regression model. In overlap weighting, treated patients are weighted by 1—PS and untreated patients are weighted by PS31). Compared with PS matching, overlap weighting keeps all patients and can enhance precision32). Compared with the inverse probability of treatment weighting, overlap weighting increases balance and precision by reducing the influence of extreme values caused by outliers33). We used standardized mean differences (SMDs) to assess the balance of covariates between the two groups. Meaningful imbalances were defined as SMDs >0.2034).
After PS overlap weighting, Cox proportional hazards regression models and robust variance estimators were used to estimate hazard ratios (HRs) with 95% confidence intervals (CIs) for outcomes. Poisson regression models were used to estimate incidence rates and incidence rate differences for the study outcomes. The cumulative incidence of reoperation with 95% CIs was calculated using the Kaplan-Meier method. Competing-risks analysis was not performed because death occurred in 0.4% (33 of 8497).
Sensitivity analyses were undertaken to test the robustness of the results. A sensitivity analysis was conducted in which the study population was restricted to those with a look-back period of at least 2 years. In addition, we performed a Cox proportional hazards model with an adjustment for clustering by institution to account for the potential influence of the treating institution. A subgroup analysis stratified LDD diagnoses into four categories, namely disc herniation, spinal stenosis, degenerative spondylolisthesis, and degenerative scoliosis. We used the hierarchical coding approach to classify the diagnoses of LDD25,35). This approach groups patients by diagnosis into a hierarchy of how controversial the use of a fusion operation is. According to previous validation studies, the sensitivity of surgical indications was 75–85% and the specificity was 80–98% compared with diagnoses provided by surgeons and medical record review25,35).
All statistical analyses were conducted using SAS version 9.4 (SAS Institute) and R 4.2.1 (R Foundation).
Between Jan 2010 and Oct 2019, 13,636 patients in the database received fusion or decompression alone for LDD. After applying the exclusion criteria, the final study cohort included 8497 patients (2051 in the fusion group and 6446 in the decompression alone group) (Fig. 1). Among the patients who received decompression alone, 2135 patients received endoscopic decompression and 4311 patients received open decompression. Patients were followed for a median of 3.3 years (interquartile range, 2.3 to 5.1), broken down to 3.1 years (2.1 to 4.5) in the fusion group and 3.5 years (2.4 to 5.3) in the decompression alone group. Mean (SD) age of the cohort was 48 (13) years, and 73% (6213 of 8497) was male. Compared with the decompression alone group, patients in the fusion group were older and had a higher proportion of patients with diabetes mellitus, osteoporosis, and high CCI. After PS overlap weighting, both groups were well balanced in all baseline characteristics (all SMDs <0.2, with most <0.1) (Table 1).
| Characteristic | Before weighting | After weightinga | ||||
|---|---|---|---|---|---|---|
| Decompression alone n = 6446 |
Fusion n = 2051 |
SMDb | Decompression alone n = 1364 |
Fusion n = 1364 |
SMDb | |
| Men, n (%) | 4912 (76) | 1301 (63) | 0.28 | 940 (69) | 940 (69) | <0.01 |
| Age in years, mean (SD) | 46 (13) | 56 (10) | 0.83 | 53 (11) | 54 (11) | 0.04 |
| Age category in years, n (%) | 0.83 | 0.13 | ||||
| 18–29 | 781 (12) | 42 (2) | 39 (3) | 40 (3) | ||
| 30–39 | 1257 (20) | 111 (5) | 117 (9) | 101 (7) | ||
| 40–49 | 1712 (27) | 328 (16) | 288 (21) | 271 (20) | ||
| 50–59 | 1608 (25) | 756 (37) | 446 (33) | 525 (39) | ||
| 60–74 | 1088 (17) | 814 (40) | 475 (35) | 427 (31) | ||
| Body mass index in kg/m2, mean (SD) | 24 (4) | 25 (4) | 0.08 | 24 (4) | 25 (4) | 0.14 |
| missing, n (%) | 2277 (35) | 666 (32) | 429 (31) | 420 (31) | ||
| Current smoker, n (%) | 1416 (22) | 414 (20) | 0.11 | 288 (21) | 299 (22) | 0.03 |
| missing | 2472 (38) | 691 (32) | 454 (33) | 440 (32) | ||
| Diabetes mellitus, n (%) | 720 (11) | 403 (20) | 0.24 | 241 (18) | 241 (18) | <0.01 |
| Osteoporosis, n (%) | 160 (3) | 212 (10) | 0.32 | 86 (6) | 86 (6) | <0.01 |
| Rheumatoid arthritis, n (%) | 73 (1) | 68 (3) | 0.15 | 31 (2) | 31 (2) | <0.01 |
| Malignancy, n (%) | 99 (2) | 64 (3) | 0.11 | 37 (3) | 37 (3) | <0.01 |
| Hemodialysis, n (%) | 20 (0) | 22 (1) | 0.09 | 10(1) | 10 (1) | <0.01 |
| Charlson comorbidity, n (%) index | 0.29 | 0.03 | ||||
| 0 | 5151 (80) | 1391 (68) | 985 (72) | 978 (72) | ||
| 1 | 768 (12) | 338 (16) | 193 (14) | 208 (15) | ||
| ≥2 | 527 (8) | 322 (16) | 186 (14) | 179 (13) | ||
| Teaching hospital, n (%) | 390 (6) | 172 (8) | 0.09 | 99 (7) | 99 (7) | <0.01 |
| Surgery year, n (%) | 0.07 | <0.01 | ||||
| 2010–2014 | 1272 (20) | 347 (17) | 236 (17) | 236 (17) | ||
| 2015–2019 | 5174 (80) | 1704 (83) | 1128 (83) | 1128 (83) | ||
a Frequency numbers were rounded to integers based on overlap weight. Weighted to adjust for age, gender, diabetes mellitus, osteoporosis, rheumatoid arthritis, malignancy, hemodialysis, Charlson Comorbidity Index, type of hospital, and surgery year.
b Standardized mean difference measured the normalized difference between the means of each covariate between the two groups.
The PS-weighted Kaplan-Meier curve showed that 3-year and 5-year cumulative reoperation rates were 6% (95% CI 5% to 7%) and 8% (6% to 10%), respectively, in the fusion group, and 7% (7% to 8%) and 10% (9% to 11%), respectively, in the decompression alone group (Fig. 2).
In the weighted cohort, there was no significant difference in reoperation incidence between the fusion group and the decompression alone group after a median follow-up of 3.3 years (weighted HR 0.85 [95% CI 0.69 to 1.04]; p = 0.11, weighted incidence rate –0.32 [95% CI, –0.90 to 0.25] per 100 person-years) (Table 2). 47% (64 of 135) in the fusion group and 30% (139 of 471) in the decompression alone group received fusion surgeries as reoperations (Table 3). Within 90 days postoperatively, there was no difference in revision rate between the two groups (weighted HR 1.17 [95% CI 0.81 to 1.70]; p = 0.40) (Table 4).
| Surgery type | Patients, n | Person-Years | Reoperation, n | Crude HR (95% CI) | Weighteda HR (95% CI) | p value |
|---|---|---|---|---|---|---|
| Fusion | 2051 | 6779 | 135 | 0.97 (0.81–1.18) | 0.85 (0.69–1.04) | 0.11 |
| Decompression alone | 6446 | 24,227 | 471 | Reference | Reference |
a Weighted to adjust for age, gender, diabetes mellitus, osteoporosis, rheumatoid arthritis, malignancy, hemodialysis, Charlson Comorbidity Index, type of hospital, and surgery year.
HR = hazard ratio; CI = confidence interval.
| Surgery type | Fusion (n = 135) | Decompression alone (n = 471) |
|---|---|---|
| Posterior discectomy, n (%) | 6 (4) | 116 (25) |
| Posterior microendoscopic discectomy, n (%) | 8 (6) | 87 (18) |
| Percutaneous discectomy, n (%) | 0 | 6 (1) |
| Anterior lumbar interbody fusion, n (%) | 7 (5) | 3 (1) |
| Posterior or posterolateral lumbar fusion, n (%) | 18 (13) | 4 (1) |
| Posterior lumbar interbody fusion, n (%) | 35 (26) | 121 (26) |
| Anterior and posterior interbody fusion, n (%) | 4 (3) | 11 (2) |
| Laminectomy, n (%) | 9 (7) | 31 (7) |
| Laminoplasty, n (%) | 17 (13) | 39 (8) |
| Microendoscopic laminectomy, n (%) | 1 (1) | 2 (0) |
| Microendoscopic laminoplasty, n (%) | 1 (1) | 8 (2) |
| Percutaneous vertebroplasty, n (%) | 1 (1) | 0 |
| Surgery for ossification of ligamentum flavum, n (%) | 1 (1) | 1 (0) |
| Debridement for surgical site infection or postoperative epidural hematoma | 27 (20) | 42 (9) |
| Surgery type | Patients, n | Person-years | Reoperation, n | Crude HR (95% CI) | Weighted* HR (95% CI) | p value |
|---|---|---|---|---|---|---|
| Fusion | 2051 | 498 | 43 | 1.09 (0.77–1.55) | 1.17 (0.81–1.70) | 0.40 |
| Decompression alone | 6446 | 1570 | 124 | Reference | Reference |
* Weighted to adjust for age, gender, diabetes mellitus, osteoporosis, rheumatoid arthritis, malignancy, hemodialysis, Charlson Comorbidity Index, type of hospital, and surgery year.
HR = hazard ratio; CI = confidence interval.
Consistent results were seen in the sensitivity analyses, in which participants were limited to those with records more than two years before the index date (weighted HR 0.88 [95% CI 0.70 to 1.11]; p = 0.28) (Supplementary Table 4) and adjusting for within-institution clustering (weighted HR 0.85 [95% CI 0.69 to 1.04]; p = 0.11). The subgroup analysis stratified by surgical indication was also consistent with the main finding; in particular, the lower reoperation risk of fusion compared with decompression alone appeared stronger in those with degenerative spondylolisthesis and scoliosis (Fig. 3).
In this large, claims-based database study of 8497 patients who had undergone fusion or decompression alone in one or two levels for LDD, we found that no significant difference was observed between the two groups.
The cumulative incidence of reoperation in our study was largely consistent with that reported in other previous studies11,18,19). A study using the Medicare database reported a reoperation rate of 9% in the decompression group and 10% in the fusion group at 3 years postoperatively19). A Swedish national registry study reported a 5-year reoperation risk of 8%18). A nationwide database study from Korea reported a 3-year reoperation risk of 6% in the decompression group and 5–8% in the fusion group11). Our results—7% in the decompression alone group and 6% in the fusion group at 3 years postoperatively—are consistent with these previous studies, albeit slightly lower11,18,19). This discrepancy might be due to our study’s participants, who were limited to age younger than 75 years old, giving a mean age (48) which was younger than in the other studies (60–76)11,18,19).
Most of the previous studies reported no difference in reoperation rates between fusion and decompression alone for LDD11–13,16–19). Our findings are in line with these previous studies. Fusion was associated with a higher risk of postoperative complications than decompression alone36); accordingly, adding debridement to the definition of reoperations might increase the relative reoperation risk of fusion. Unlike other claims-based database studies11,12,19), we limited our patients to those who underwent fusion or decompression alone in one or two levels, similarly to major randomized controlled trials on this topic16,17,36). Patients who receive fusion in multiple levels have much higher reoperation risk than those who do so in one or two levels37). Thus, including these patients in the fusion group would increase the reoperation risk following fusion.
This study had several limitations. First, our findings might have been affected by unmeasured confounders, such as surgeon experience, and radiologic information, such as the degree of instability, alignment, or degeneration. Furthermore, detailed clinical information such as surgical indications, diagnoses, and physical findings are not available in our database. These unmeasured confounders could also have biased our results. Nevertheless, use of PS overlap weighting provided a good balance for all measured confounders. Furthermore, the fact that high-risk patients (ex. degenerative scoliosis and degenerative spondylolisthesis) are more likely to receive fusion than decompression alone may have led to selection bias, with increased reoperation incidence in the fusion group and a skewing of results toward the null12). In addition, our subgroup analysis stratified by surgical indication was consistent with our main findings. Second, our study cohort included a variety of lumbar pathogenesis, which raised concern for heterogeneity. We did not adjust for surgical indications because this factor strongly influenced whether to perform fusion or decompression alone and could be an instrumental variable. If this factor was added to PS model, it would have increased bias38). Furthermore, the results of our subgroup analysis stratified by surgical indications were consistent with our main results. Third, our median follow-up of 3.3 years may not have been sufficient to investigate the reoperation risk accurately10,11,19). A large prospective cohort study with at least 10 years’ follow-up duration is warranted to confirm our findings. Fourth, inaccuracies in claims-based databases, such as coding errors, could lead to measurement bias. Nevertheless, a previous validation study reported that claims data accurately reflected information about surgical procedures25). In addition, we could not find out what caused reoperation. As the causes of reoperation differed between fusion and decompression alone, we should be careful about the interpretation of our results. Furthermore, the reoperation rate included any type of surgery, and thus may have differed from that at the index level. In practical terms, however, this information is likely to be as meaningful to patients as the reoperation rate at the same level. In addition, it appears almost impossible to deny the possibility that primary lumbar surgery may have caused degeneration at the adjacent level. Fifth, while we set a one-year look-back period, it is almost impossible to ensure that the index surgery for LDD was the primary one. However, our sensitivity analysis restricting patients to those with a 2-year look-back period demonstrated consistent results with our main results. Thus, we do not consider that this substantially influenced the results. Fifth, our results may have limited generalizability regarding age because the mean age was 48 years. Furthermore, our findings reflect the results of real-world clinical practice in Japan; given that indications for fusion and decompression alone vary from country to country4,6,9,11), it is unclear whether our results are extrapolatable to other countries. Lastly, we did not have data on patient-reported outcomes. Moreover, we could not assess length of stay, readmission, or cost because we could not determine what types of diseases were responsible for additional treatments and/or readmission. A future cost-effectiveness study is needed to evaluate the effectiveness of fusion surgery to mitigate the risk of reoperation.
In conclusion, this study suggested that no significant difference was observed between the fusion and decompression groups. Our findings should make surgeons consider the indications for fusion or decompression by weighing its reoperation risk.
The authors disclose receipt of the following financial or material support for the research, authorship, and/or publication of this article: Koji Kawakami has received research funds from Eisai Co., Ltd., Kyowa Kirin Co., Ltd., Sumitomo Pharma Co., Ltd., Mitsubishi Corporation, and Real World Data Co., Ltd.; consulting fees from LEBER Inc., JMDC Inc., Shin Nippon Biomedical Laboratories Ltd., and Advanced Medical Care Inc.; executive compensation from Cancer Intelligence Care Systems, Inc.; and honoraria from Mitsubishi Corporation, Pharma Business Academy, and Toppan Inc. All other authors received no financial or material support for the research, authorship, and/or publication of this article.
We thank Dr. Guy Harris DO of DMC Corp (www.dmed.co.jp) for his support with the writing of the manuscript.
