Long-term Reoperations After Lumbar Spinal Stenosis Surgery

The use of fusion surgery in addition to decompression in the treatment of degenerative lumbar spondylolisthesis (DS) with spinal stenosis (LSS) is a long-standing controversy in spine surgery. The main goal of LSS surgery is to decompress the nerve roots. In the 1990s, two observational studies recommended that decompression with additional fusion should be the treatment of choice. Consequently, the practice shifted towards more complex fusion procedures. In 2016, Swedish and one American randomized controlled trials (RCTs), one Swedish and one American, reported conflicting data about reoperations. The study from the US showed a higher frequency of reoperations after decompression only compared to the decompression and fusion, while the Swedish study showed no difference. In 2020, Austevoll et. al. published a study from the Norwegian registry for spine surgery (NORspine) comparing relative effectiveness micro-decompression alone vs decompression and instrumented fusion (5). This study had a non-inferiority design, similar to the RCT published by the same author in 2021 (6). Both studies concluded that micro-decompression alone was non-inferior to decompression and instrumented fusion regarding clinical outcomes and reoperation rates. Despite an increasing number of studies showing no extra benefits from the more risky and complex additional fusion procedures, the surgical practice has changed little outside Scandinavia. This is probably due to the concerns about subsequent instability and higher long-term reoperation rates among those operated with decompression only.

This study is a long-term follow-up of the NORspine relative effectiveness study from 2020, comprising 794 patients having an index operation for LSS and DS between September 19th 2007 and December 21st 2015 (index study). The present study aims to assess long-term reoperations resulting from everyday clinical practice. The investigators alternative hypothesis is that micro-decompression alone is non-inferior to decompression plus instrumented fusion. The non-inferior margin is specified to correspond to a number needed to treat = 8 to avoid reoperation on one patient in the first 10 postoperative years, corresponding to a between-group difference of 12.5 percentage points (100/8 = 12.5).

The dataset will contain NORspine data at baseline, 3 months, one year and long-term follow-up performed at 7 to 15 years (NCT03469791). NORspine also provides dates of death for all deceased patients. The Norwegian Patient Registry (NPR) will serve as an external data source, providing data about all reoperations until August 31st 2023 and the baseline Charlson Comorbidity Index (CCI). NPR data will ensure follow-up of 8 to 16 years concerning reoperations. For these cases, a review the electronic health records to validate NPR data regarding classification of reoperation (level, indication, surgical technique, total number of reoperations, spinal cord stimulation, reoperations within 90 days after the index operation, participation in other studies).

In the index study propensity score matching (PSM) was performed to reduce the risk of selection bias due to case-mix. The same PSM will be used in the present study. matching. The following parameters were included in the calculation of the propensity score: Age; Gender; American Society of Anaesthesiologists (ASA) grade; Body Mass Index (BMI); Smoking, Oswestry Disability Index (ODI), Numeric Rating Scale for leg pain and back pain, Euroqol 5D (EQ-5D-3L), foraminal stenosis, degenerative disc disease, predominating back pain, number of levels operated on and neurological palsy. The propensity scores were derived from a logistic regression model and reflected a patient's theoretical baseline probability for being instrumentally fused. Using the '1:1 matching without replacement' method, pairs of fused and non-fused patients with a difference in propensity scores less than 0.2 in the logit of the standard deviation were formed. Statistical Package for the Social Sciences (SPSS) version 24 was used for propensity score matching. For comparison, complete case analyses will also be performed.

In this non-inferiority designed study, the primary outcome will be the occurrence of a reoperation, defined as new lumbar spine operation more than 90 days after the index operation until August 31st 2023 (range: from 8 to 16 years after the primary operation. The rate of reoperation within 90 days (n%) of will be reported separately, and will be classified as a complication to the index operation. Survival analysis will be used to detect whether the probability for reoperation during the first 10 years is more than 12.5 % higher (non-inferiority margin) in the micro-decompression group than in the decompression and instrumented fusion group or not. This is to be tested by deriving a 90 % confidence interval (CI) for the difference of the survival function (S(t)) at 10 years between the groups (S(t) in decompression and instrumented fusion group - S(t) in micro-decompression group) where t = 10 years. Thus, a lower limit of the CI above -0.125 will indicate that, in an everyday clinical practice setting, micro-decompression would be non-inferior to decompression and instrumented fusion regarding the risk of reoperation. In addition, the investigators will report survival functions and observed reoperation rates at multiple timepoints (2, 5 and 10 years), hazard ratio (HR) for reoperation (at 2, 5 and 10 years, as well as complications (including reoperations within 90 days). The indication for reoperation, surgical techniques and risk factors associated with reoperation will also be investigated.

For the power analysis conducted with PASS 2019, two assumptions were made. 1) Expected reoperation rate: Based on NORspine data, about 27.1 % of lumbar surgical procedures are performed on patients previously operated. Using this as a proxy for the expected probability of reoperation in the total cohort. 2) HR = 1.6 (decompression and instrumented fusion) approximates the event ratio between the groups. Applying these assumptions, choosing a type 1 error = 0.05 and power 0.8 gives sample size = 269 per group. The index study had a sample size of 285 in each PSM group. Since the present study will use national administrative registries and patient's health records, substantial loss to follow-up is not expected.

For statistical analyses SAS Enterprise Guide 8.3 will be used, including descriptive statistics, tests for data distribution, cross-tabulations with χ2 test, Student t-tests, and Mann-Whitney U tests. Reoperation survival, hazard and rates will be assessed by stratified log-rank test, cox regression modelling (time-dependent if the proportional hazard assumption is violated), Kaplan-Meier plots and multivariable regression analyses.