The NORDSTEN Study - Degenerative Spondylolisthesis

Purpose:

The aim of the study is to compare the efficacy of two principally different surgical methods for treating LDS and to investigate whether predictors will be associated with an intervention-related difference in outcome between the two groups.

Methods:

The Oswestry Disability Index (ODI) will be used as primary outcome measure in analyses at 2 year, 5 year and 10 year follow-up. Depending on whether the participants obtain a 30% or more ODI- improvement from before operation to follow-up, they will be dichotomized in a success group (responders) and in a non-success group (non- responders). The primary question is whether the proportion of patients with successful outcome in the DA group is not markedly less than in the DF group, at 2 years, 5 years and 10 years, respectively. Investigators have predefined the margin of "markedly less" to be δ = 0.15 (15%) which is consistent with the between- group difference Blumenthal et al used in the power analysis in a similar Food and Drug Administration- regulated non- inferiority RCT. With this margin it will be necessary to treat 6.7 patients with fusion in addition to decompression in order to obtain one patient with successful outcome.

The null hypothesis is that the proportion of patients with a successful outcome in the DF group (nDF) is higher than the proportion in the DA group (nDA) by an amount of at least 0.15 (δ):

H0: nDF >= nDA + δ

H0 will be tested by forming a 95% CI for the difference of proportions and H0 will be rejected if the upper limit of the confidence interval (CI) is less than 0.15.

By rejecting H0, the alternative hypothesis will be accepted:

H1: nDF < nDA + δ

By accepting H1, the conclusion will be that DA is non-inferior DF and hence as good as DF.

In addition to the efficacy analyses (main analyses)we aim to investigate the following:

1. Whether baseline parameters are associated with clinical outcomes,and whether treatment effect modifiers can be identified at the defined follow-ups (at 2 years, 5 years and 10 years) from the following list of variables:
2. Treatment group
3. Patient age
4. Gender
5. Comorbidity (ASA group)
6. Body Mass Index
7. Smoking
8. Preoperative ODI score
9. Preoperative NRS back pain
10. Preoperative NRS leg pain
11. Hopkins symptom check list (HSCL-25)
12. The magnitude of olistheses and local kyphosis measured on standing x-rays
13. Segmental instability verified on flexion/extension x-rays
14. Presence of foraminal stenosis in the level of olisthesis
15. Orientation of the facet joint in the level of olisthesis
16. Amount of facet joint fluid in the level of olisthesis
17. Degree of disc degeneration in the level of olisthesis
18. Disc height in the level of olisthesis
19. Pelvic incidence

II. Whether the surgeon's preference for treatment before randomization is associated with clinical outcomes at follow-up (2-year, 5-year and 10 year). In this study, we will use the similar primary and secondary outcome measures as for the efficacy study.

III. Health-economic comparison of decompression alone and decompression with fusion at 2-, 5- and 10-year follow-up. Details for the analysis of data up to 5-year follow-up are provided in the uploaded "Statistical Analysis Plan, NORDSTEN-DS"

This will be published in separate articles.

Sample size:

The sample size calculation is based on the hypothesis that the 2-year results for the DA group is as least as good as for the DF group when comparing the proportions of responders in each group. The sample size is computing by using the Blackwelder methodology. Based on data from the Norwegian Spine Register, the proportion of responders for the whole treatment group is expected to be 0.70. Choosing a type 1 error = 0.05, power = 0.80 and δ = 0.15 gives a sample size of 116. Considering these assumptions and adding 10% for possible dropouts, a total of 128 patients are required in each group. Using Multiple Imputation for missing data, we consider power acceptable even with some higher dropouts at 5- and 10-year follow-up.

Protocol deviations

The patients have major deviations from protocol if they:

Have not received operative treatment in accordance with randomized allocation; Have received operative treatment in accordance with randomized allocation and operated with a new lymbar operation during the follow-up period; Have not provided informed consent; Have withdrew the informed consent and claimed their data withdrawn from analyses.

According to deviations to protocol the following analysis sets are defined:

Modified intention-to-treat (mITT) set (previously named as "Full Analysis Set): all randomised patients with primary operation according to the randomly assigned study treatment and with data on the primary outcome variable (ODI) at one or more time point.

Per Protocol Set (PPS): All randomized patients without major deviations from protocol and with data on the primary outcome variable at baseline and 10-year follow-up.

Statistical analysis:

To determine whether randomization is successful descriptive statistics will be used for comparing baseline characteristics between the groups.

Differences in responder rates for ODI, ZCQ and NRS for leg/back pain(proportions of patients dicothomized into 'success'), including the Newcombe hybrid score confidence interval, will be estimated and tested according to non-inferiority. If there are patients with missing data at follow-up, Multiple imputation (MI) will be performed including plausible baseline data and follow-up data as explanatory variables in the imputation model.

Update (amndment to original protocol) recorded in the Statistical Analysis Plan for the 10-year follw-up: Following the original power calculation, the trial requires 232 patients (116 per group) to achieve 80% power to ensure that the lower bound of a two-sided 95% confidence interval excludes a 15-percentage-point higher rate of clinically important improvement in ODI in the fusion group. At the five-year follow-up, 189 patients were included in the per-protocol set (PPS) analysis. Due to an an expected further decrease in the number of participants meeting the PPS criteria by the 10-year follow-up and following discussions within the study group and with the responsible biostatistician, we have decided that the PPS analysis will no longer be used as decision criteria for claiming non-inferiority of decompression alone. Instead, results for the primary outcome in the per-protocol cohort will be presented descriptively for each treatment group, without statistical hypothesis testing for non-inferiority. Consequently, the assessment of relative efficacy will rely exclusively on the modified intention-to-treat set utilizing multiple imputation of missing 10-year ODI scores.

For similar reason-likely limited statistical power due to even lower follow-up at 10 years (n=226 at 5 years)-the results for the primary outcome in the modified intention-to-treat complete-case set (originally planned as a sensitivity analysis) will be presented descriptively only, without formal between-group hypothesis testing. We will retain the originally planned sensitivity analysis using the modified intention-to-treat set in which missing 10-year ODI values are replaced with the observed 5-year ODI score when available. This analysis will include at least 226 participants, corresponding to the number of complete cases at the 5-year follow-up.

Categorical secondary outcomes will be analyzed with Fisher mid-P tests and Newcombe hybrid score intervals. The GPE responses will be analyzed with a proportional odds logistic regression model.

It will be used a linear mixed model analysis to evaluate the continous secondary outcomes, and to evaluate the effect of each efficacy variable over time and between groups (all follow-up measurements from inclusion to follow-ups will be included). In the mixed model patients are not excluded from the analysis of an efficacy variable if the variable is missing at some, but not all time points after baseline.

A significance level of 5% will be used throughout.

Data will be inaccessible to the research group until all available 10-year follow-up participants has completed the 10-year questionnaires.The data collected between 5- and 10-year follow up will be inaccessible to the research group until all available 10-year follow-up participants has completed the 10-year questionnaire. Faculty of Research support, University of Oslo will declare this inaccessibleness and state which date the data has been accessible for the investigators.

Interim analysis and stopping rules:

When 75 patients in each group have completed the 12 months follow- up, the investigators will do an interim analysis. Due to ethical considerations the inclusion of patients will be terminated if the analysis reveals one of the following:

1. The proportion of patients with successful outcome in the DF group is higher than in the DA group by an amount of 0.20.
2. The proportion of patients needing reoperation due to any condition in the operated level(s) is statistically significant higher in one of the groups.

The interim analysis was conducted on February 28, 2017 by an independent statistician blinded for treatment adherence. Only data on reoperations and on the primary outcome measure (ODI) was available to the statistician. None of the stop criteria were fulfilled. The statistician informed the steering committee, via the central coordinator, that the study could be continued. Further information about the analysis was not disclosed.

The reporting of the trial will be based on an adapted Consolidated Standards of Reporting Trials (CONSORT) checklist for reporting non- inferiority trials.