We used data from the inception cohort of the Swiss Spinal Cord Injury (SwiSCI) study [18]. SwiSCI study is a cohort established as a collaboration among four major rehabilitation centers across Switzerland (Swiss Paraplegic Centre, Nottwil; Klinik für Neurorehabilitation und Paraplegiologie-REHAB Basel, Basel; Clinique romande de readaptation, Sion; and Balgrist University Hospital, Balgrist) which serve as regional catchment areas for individuals requiring specialized care post-injury. SwiSCI Inception Cohort prospectively enrolled individuals with SCI who were admitted for inpatient rehabilitation in one of its participating centers in Switzerland. Data were collected in the study centers at four time points following the date of SCI diagnosis: at 28 days (range 16–40 days, T1), 84 days (70–98 days, T2), 168 days (150–186 days, T3), and at discharge (10–0 days before discharge, T4). Our analyses focused on admission to rehabilitation (T1), which represents the study baseline, and rehabilitation discharge (T4). Data are collected by extraction of routine clinical information from the medical records, by clinical assessments, and by paper-and-pencil questionnaires. A comprehensive list of commonly utilized metrics within the collaborating centers was developed, with a focus on prioritizing and standardizing established measures across all four centers. The SwiSCI Inception cohort data model is based on the International Classification of Functioning, Disability and Health (ICF), and the Brief ICF SCI Core Sets in the early post-acute context was used as a reference for the clinical setting. Additionally, whenever applicable and accessible, preference was given to incorporating the “International SCI Basic Data Sets” recommended by the International Spinal Cord Injury Society (ISCOS) (https://www.iscos.org.uk/international-sci-data-sets). Detailed information on the study design and collected data have been reported elsewhere [18, 19].

We enrolled all adults (≥18 years old) from May 2013 to September 2020, who were admitted to any of the four participating rehabilitation centers. Individuals with an SCI attributable to a congenital condition, neurodegenerative disorder, or Guillain–Barré syndrome, or who had a new SCI in the context of palliative care, were excluded from the study. Furthermore, individuals with SCI who had malignant neoplasms or those in palliative/end-of life care were excluded. Finally, we excluded those with previous history of CVD to create a homogenous baseline cardiovascular risk profile of our analysis population. This is also in accordance to how most studies in cardiovascular risk profiling were conducted in the literature.

The SCI characteristics included SCI lesion etiology (e.g., traumatic vs. non-traumatic, causes of the injury), level and completeness of the injury (motor complete and incomplete), and the pattern of NTSCI injury onset (including acute, sub-acute, prolonged). The level of injury was classified as tetraplegia (at level C2-C7) and paraplegia (level T1-S5), and the completeness of injury into complete motor injury (AIS A and B) and incomplete (AIS C and D) based on the International Standards for Neurological Classification of Spinal Cord Injury (ISNCSCI) [20]. In addition, time since injury and duration of rehabilitation at the SwiSCI rehabilitation center were derived from medical records. Further, demographic characteristics such as age at baseline, sex, information on comorbidities and medication use were obtained from the SwiSCI database and were derived from patient’s medical records.

Venous blood samples were obtained from each participant after an overnight fast. Samples were then sent to respective hospital laboratories for lipid and glucose profiles. Waist circumference (WC) was measured after bowel care. Measurement was taken at the end of a normal exhale, between the lower rib and the top of the hip bone. A flexible tape measure with a precision of 0.5 cm was used. Weight was measured using an electric wheelchair scale. The wheelchair’s weight was subtracted from the total weight of the subject with the wheelchair to determine the subject’s weight expressed in kilograms (kg). Body mass index (BMI) was computed employing the standard formula [weight in kilograms/(height in meters)2].

We identified individuals with CMD using the criteria provided by the SCI-specific clinical guideline [21]. CMD factors included blood pressure, fasting lipid profile, fasting glucose, and anthropometric measures, that was also used individually for longitudinal modeling. The risk of developing the first cardiovascular event within the next 10 years was assessed using the Framingham risk score (FRS) [22]. The FRS of each study participant was computed at discharge from initial rehabilitation stay using the following variables: (a) age, (b) sex, (c) systolic blood pressure (SBP), (d) total cholesterol (mg/dL), (e) high-density lipoprotein cholesterol (mg/dL), (f) diabetes, and (g) current smoking [22].

We summarized continuous variables using median and interquartile range (IQR) as prescribed by the International Spinal Cord Society (ISCOS) Standards of Data Analysis and Reporting [23]. We log-transformed all non-normally distributed continuous variables. Categorical variables were presented as numbers and percentages. To compare the differences in demographic characteristics, injury characteristics, clinical parameters, lifestyle factors, and comorbidities at baseline between TSCI and NTSCI, we used Wilcoxon signed rank test and chi-square test, as appropriate.

We used a paired t-test to compute the longitudinal changes in cardiometabolic parameters from beginning to end of rehabilitation for individuals with TSCI and NTSCI. We also used a multilevel mixed model using random slope of each individual trajectory by residual maximum likelihood estimation. The longitudinal model was adjusted for age, sex, smoking history, alcohol use, time since injury, prevalent and incident CMD, injury completeness and injury level. Furthermore, we included an interaction term (injury etiology and rehabilitation time) to account for time specific changes in CVD risk factors.

According to the level of injury, we explored the longitudinal changes in cardiovascular risk of the study participants. We similarly used multilevel mixed model using random intercept and individuals as clusters. We used anthropometric measures, blood pressure, fasting lipid profile, and fasting glucose as outcome variable. We used injury etiology (TSCI versus NTSCI) as our predictor variable, with similar model adjustments as previously mentioned.

Finally, we investigated on risk factors for changes in the components of CMD. For this, we performed multivariable linear regression using discharge values (anthropometric measures, blood pressure, fasting lipid profile, and fasting glucose), that were fitted among individuals with TSCI and NTSCI separately. Model adjustments were done as previously mentioned. This was done to explore the longitudinal association between age, sex, injury severity, rehabilitation duration and lifestyle factors and CMD factors.

All statistical analyses were performed using the Stata 16.1 (StataCorp LLC, College Station, TX) for Windows. All computations were done using two-tailed tests, and a p-value of < 0.05 was considered statistically significant.

We performed sex-stratified analyses to determine sex-specific associations. We performed age-stratified analysis to determine age-specific associations based on median population age (55 years). To detect selection bias, we also compared the excluded and the included study cohort. For missing data, we tabulated missing exposures and outcomes, and performed list-wise deletion in our regression analyses.