Patient Reported Recovery of Muscle Strength After Surgery for Cervical Radiculopathy, Impact of Duration and Severity on Recovery.

The most common cause of cervical operation is Degenerative cervical radiculopathy (DCR), i.e. a neural impingement by disc herniation or spondylotic changes. While pain is often the most prevalent indication of surgery, motor weakness is frequently reported and is associated with less favourable outcome.

Preoperative motor weakness is reported up to 70%. Some studies suggest that both the severity and duration of preoperative motor weakness are associated with persistent postoperative weakness, implying that delayed intervention may reduce the likelihood of full neurological recovery.

Moreover, current evidence does not establish whether surgery is preferable to conservative treatment in patients with motor weakness. The optimal timing of surgery and the grades of weakness most likely to benefit also remain unclear. A systematic review was unable to draw firm conclusions because studies varied substantially in strength assessment methods, outcome definitions, and follow-up duration.

Thomé et al. reported that, in lumbar radiculopathy, the association between early surgery and motor recovery was dependent on both baseline muscle grade and symptom duration.12 However, these findings may not be generalisable to cervical radiculopathy given key anatomical and pathological differences.

This study aims to evaluate the association between the preoperative duration and baseline severity (muscle strength grade) of motor weakness and complete strength recovery at 12 months after surgery for DCR. As a secondary, exploratory aim, we will assess whether there is a severity grade and/or duration window in which the probability of complete recovery is highest.

Primary Objective Evaluate the association of (1) patient-reported duration of weakness prior to surgery (timing proxy) and (2) baseline objective motor deficit severity (MRC grade) with patient-reported complete strength recovery (Yes/No) at 12 months following single-level surgery for cervical radiculopathy.

Exploratory objectives

• Identify candidate cut points for timing and/or baseline MRC using tree-based recursive partitioning to predict 12-month complete recovery (cervical-specific threshold discovery).
• Evaluate predictive discrimination using ROC/AUC for timing, MRC, and combined models (with internal validation if feasible).

Secondary Objectives

• Assess effect modification (Timing × MRC interaction) for the primary endpoint.
• Evaluate associations of timing and baseline MRC with NDI and NRS outcomes (arm pain, neck pain, headache) at 12 months (and at 3 months if available). -Evaluate postoperative strength trajectory using a binary outcome measure complete recovery (yes/no) at 3 and 12 months with random effects (Thomé-analogue trajectory analysis under registry constraints).
• Perform etiology-stratified subgroup analyses (disc herniation vs spondylotic foraminal stenosis) to assess heterogeneity of associations.

Hypothesis Our null hypothesis is that the prognosis of recovery for DCR patients with paresis is independent of paresis severity, duration and underlying pathology. Our alternative hypothesis is that the recovery rate of paresis after DCR surgery is less for more severe paresis of longer duration and will vary according to underlying pathology.

Study populations The study will include consecutive patients operated for DCR and registered in NORspine form January 2012 through 2022, approximately 12.000 cases. Patients operated for cervical myelopathy will be excluded. Otherwise patients under the age of 16, those operated for fractures, tumours and primary infections, as well as those unable to give informed consent, are not included in the NORspine.

Data sources The project will use prospectively collected data from The Norwegian registry for spine surgery NORspine as the only data source, The NORspine is a comprehensive national clinical registry, with a capture rate of patients operated for DCR of 80% from in 2021 and a response rate at one year follow up (2022) of 80%. Coverage at the institutional level is 95-100%. A wide range of established and possible prognostic factors are recorded in the NORspine prior to surgery. Patient reported outcome measures (PROMs), are recorded at admission for surgery (baseline) and 3 and 12 months (maximum follow up time) after surgery.

Predictors / Exposure Variables

-Timing proxy: patient-reported weakness duration prior to surgery (registry item; English translation) Baseline question: "Do you have reduced strength (weakness) in the affected shoulder, arm, or hand?" (Yes/No).

If Yes: duration (choose one; optional numeric field where available):

• Less than 24 hours (optional: number of hours)
• Less than one week (optional: number of days)
• 1 week to 3 months (optional: number of weeks)
• 3 months to 12 months
• More than 12 months

Operationalization (prespecified):

• Primary: derive duration in days when numeric detail is available (hours/days/weeks converted to days).
• Otherwise: treat duration as an ordered categorical predictor using the registry intervals.
• Sensitivity: apply a prespecified category-to-days mapping (e.g., midpoint assumptions) when numeric detail is missing; assumptions reported explicitly.

Baseline objective motor deficit severity (MRC) Preoperative MRC grade (0-5) recorded by clinician/surgeon for a prespecified key muscle group corresponding to the index level, according to the institution's standard mapping. The cohort is restricted to objective weakness at baseline (MRC < 5).

Planned MRC coding (depending on data)

• Primary coding: severe 0-2 /moderate 3 / mild 4
• Secondary coding: severe ≤3 vs non-severe ≥4. Note: Tree-based methods may identify alternative splits; these will be reported as exploratory and will not redefine confirmatory coding.

Covariates / confounding control Potential confounders/covariates will be evaluated a priori based on clinical knowledge and registry availability by directed acyclic graph (DAG). Adjusted models will be specified before analysis and will not rely on automated or purely significance-driven variable selection.

Statistical Analysis Plan

Primary Confirmatory Analysis (Prediction of 12-month complete recovery)

• Outcome: complete recovery at 12 months (Yes/No).
• Model: multivariable logistic regression.
• Key predictors: weakness duration (days with non-linearity assessment; otherwise ordinal categories) and baseline MRC severity (0-3 vs 4; secondary 0-2/3/4 if feasible).
• Confounding control: covariates will be evaluated and specified a priori; models will not rely on automated or purely significance-driven variable selection.
• Effect modification: prespecified duration × MRC interaction; etiology interaction and/or stratified reporting.

Secondary Trajectory Analysis (binary longitudinal strength outcome at 3 and 12 months)

• Outcome: Complete recovery (yes/no) or complete or improved recovery (yes/no) status at 3 and 12 months
• Model: logistic mixed-effects model with patient random intercept and fixed effect of time (12 vs 3 months), plus baseline predictors and prespecified interactions.

Exploratory Threshold Discovery (Tree-based)

• Conditional inference tree / recursive partitioning to identify candidate cervical-specific cut points in duration and/or baseline MRC predicting 12-month complete recovery.
• Internal validation via cross-validation and/or bootstrap where feasible.
• Tree-based methods are used only for exploratory threshold discovery and will not be used for confounder selection.

Exploratory Discrimination (ROC/AUC)

•ROC/AUC for models using (a) timing alone, (b) MRC alone, and (c) combined predictors, with internal validation if feasible.

Missing Data

• Describe extent/patterns of missingness in predictors and outcomes.
• Primary analysis: complete-case for required 12-month outcome; sensitivity analyses as appropriate.
• Longitudinal models incorporate partially observed 3/12-month outcomes under MAR conditional on observed covariates and earlier outcomes; sensitivity analyses may include inverse probability weighting and/or multiple imputation.