This prospective, multi-centre comparative study was reported in accordance with the Strengthening the Reporting of Observational studies in Epidemiology (STROBE) statement [9]. Patient inclusion started in August 2012 and the last patient was enrolled in February 2021. Data was collected at baseline, 6, 12, 26, 38, 52 and 104 weeks after inclusion through questionnaires. The study protocol was approved by the Medical Ethics Committee of the Leiden University Medical Centre entre was independently obtained prior to patient inclusion and written informed consent was secured from all patients. In addition, written informed consent was secured from all patients. The study was registered July 3, 2012 in the Netherlands Trial.

Patients aged 18 to 75 years old with CRS due to a herniated disc, experiencing disabling symptoms or tingling for at least two months, were eligible. Diagnosis of cervical radiculopathy was confirmed by a neurologist or neurosurgeon in participating hospitals. Eligibility required VAS pain ≥ 40 and MRI confirmation of herniated disc with root compression. Exclusion criteria were arm paresis (MRC < 4), myelopathy, prior cervical surgery, instability requiring fixation, pregnancy, insufficient Dutch language skills, planned emigration or participation in another trial. After presentation to the neurosurgeon conservative and surgical treatment were discussed and in agreement a decision was made.

The patient was positioned prone with the neck slightly extended under general anaesthesia. The affected cervical disc level was identified with the help of fluoroscopy. A small transverse incision was made on the side chosen by the surgeon.

The pre-vertebral space was accessed medial to the carotid sheath and the anterior cervical spine was exposed. Caspar spreader and 2 distraction pins were placed in the affected segment. Care was taken not to damage the adjacent level discs. A standard anterior discectomy with the aid of loupe magnification or microscope (depending on the surgeon’s preference) was performed in all cases. The posterior longitudinal ligament was opened and the nerve root and dura were decompressed adequately. A PEEK cage was placed in the intervertebral space. If required, a vacuum drain was placed and the wound was closed in layers. Postoperative care followed standard protocol with a hospital stay of one- to two-days.

Treatment for CRS aims at alleviating pain and the preservation or restoration of ordinary daily activities. Several studies have shown that providing clear information about the patient’s condition and prognosis, along with dependable counselling, can significantly reduce patient anxiety and uncertainty [10, 11]. Patients received standard care, which included education on the favourable prognosis of CRS, encouragement to resume normal activities, and a stepwise approach to pain management with paracetamol, NSAIDs, or tramadol if necessary. Normally, soft collars or physiotherapy are not prescribed, however patients were permitted to initiate these treatments. A schedule for graded activities is established together with the patient in advance.

In cases of progressive neurological deficits or intolerable pain, the general practitioner or neurologist referred the patient back to the neurosurgeon to explore potential surgical options.

Baseline characteristics, including age, gender, weight, length, personality type D Scale-14 (DS-14) and smoking status, were retrieved. Questionnaires evaluating outcome measures were sent to the patients at baseline, 6, 12, 26, 38, 52 weeks and 104 weeks after enrolment. Primary outcomes included the Visual Analogue Scale for arm pain (VAS arm) and the Neck Disability Index (NDI). The VAS arm score ranges from 0 mm (no pain) to 100 mm (worst pain). The NDI, assesses pain, work-related activity and daily activities, with scores ranging from 0 (best) to 50 (worst), converted to a 100-point scale. Both measures are recognized for its reliability and validity in evaluating patients with cervical pathology [12,13,14,15,16]. The DS-14 is a Dutch questionnaire assessing ‘Distressed-personality’. The scale ranges from 0 to 28, with higher scores reflecting higher levels of negative affect [17, 18]. Secondary outcomes were the VAS neck pain and the EuroQol VAS (EQ-VAS). The EQ-VAS is a self-reported assessment tool for overall health, ranging from 0 (worst imaginable health) to 100 (best imaginable health).

Clinical trials often assess treatment effectiveness, however statistical improvements may not always reflect patient benefit. Thus, to determine whether a treatment offers real benefits, the concept of Minimal Clinically Important Difference (MCID) is used to evaluate its effectiveness [19]. For the VAS arm and neck pain, a 30% decline deemed a valuable benchmark for identifying clinically meaningful improvement [20]. For the NDI the MCID is defined as 20 points on a 100 point scale [19, 21, 22].

Another evaluation method to dichotomize outcome data to evaluate the percentage of patients that reports a ‘good outcome’, is to use pre-published cut off values for ‘success’. The following cut off values were used and considered as a ‘good outcome’: VAS Arm pain ≤ 25, NDI ≤ 24, VAS neck pain ≤ 35 and EQ-VAS ≥ 69 [23].

To address potential sources of bias we have done a consecutive evaluation of trial candidates.

Originally, the study started as a Randomized Controlled Trial (RCT) [24]. The target sample size was set to 200 patients in both arms (S1 Appendix). During patient recruitment, inclusion appeared challenging. After including 12 patients within the RCT format, the study was converted to an observational cohort study design. After nine years, the achieved participant count reached 141 patients. The need for information on this topic persists, which is why we decided to initiate analysis after including 141 patients.

Differences between groups at all follow-up points were analysed using linear mixed-effects models including a random intercept for each subject to account for repeated measurements. Data from participants with incomplete VAS arm pain scores were included, multiple imputation was deemed unnecessary for mixed-model analysis of longitudinal data [25]. The analysis, stratified by treatment group, included overall group comparisons adjusted for baseline differences of the clinical parameter. Secondly, we added the covariates age, gender, smoking behaviour, BMI and DS-14.

Data management was conducted with the ProMISe system of the Department of Medical Statistics and BioInformatics of the Leids University Medical Center.

IBM SPSS software, version 27.0, was used for all statistical analysis.