The design of the study was retrospective and observational. Informed consent was obtained from the patient/next of kin before the execution of the MRI. Our institutional review board approved the present study with a waiver of informed consent, due to retrospective observational design (Ospedale Pediatrico Bambino Gesù; protocol number 1867/2019). The study was conducted in accordance with the Declaration of Helsinki.
Patients were retrospectively recruited by reviewing the imaging archive of our institution from 2011 to 2020 with the following inclusion criteria: (a) presence of MRI exam including spinal cord examination with axial and/or sagittal T2 images; (b) genetic confirmation of KSS disease obtained with mitochondrial DNA analysis from muscle biopsy, blood, or urine samples. The only exclusion criterion was the presence of significant image artifacts.
Patients underwent multiple MRI to assess the central nervous system (CNS) disease progression according to clinical indications. MRI was performed for each patient on the same 3 T magnet (Siemens Skyra, Siemens Medical Systems, Erlangen, Germany), with a protocol including at least axial T2-weighted (TR = 8600, TE = 122, slice thickness = 3 mm); axial FLAIR (TR = 9000, TE = 85, slice thickness = 3 mm); axial T1-weighted (TR = 500, TE = 9.9, slice thickness = 3 mm); DWI (TR = 9000, TE = 98, slice thickness = 3 mm, b value = 0/1000); axial and sagittal T2-weighted images for the spinal cord, including cervical, dorsal, and lumbar segments (TR = 4000, TE = 122, slice thickness = 2.3 mm).
Spinal cord involvement was evaluated on sagittal and axial T2-weighted images. For statistical correlation, spinal involvement was defined as a binary variable (0/1). The severity of spinal cord disease was classified in three different grades: no involvement (grade 0), focal involvement (limited to one spinal segment, grade 1), or extensive involvement (two or more spinal segments, grade 2) (Table 1). The pattern of involvement was qualitatively evaluated on axial T2-weighted images.
To compare spine and brain disease, we calculated a radiologic score on axial T2 images, based on typical disease locations described in previous works [5, 7,8,9]. The score had a maximum of 16 points, distributed as follows: subcortical white matter (1 point), deep white matter (1 point), basal ganglia (2 points for the involvement of > 2 among putamen, globus pallidus, and caudate; 1 point if less), thalamus (1 point), hypothalamus (1 point), mesencephalon (2 points for plate and tectum involvement; 1 point if less), pons (2 points for plate and tectum involvement; 1 point if less), medulla (2 points for plate and tectum involvement; 1 point if less), cerebellar white matter (1 point), cerebral atrophy (1 point), and cerebellar atrophy (1 point) (Table 1).
Images were evaluated separately by two research fellows in pediatric neuroradiology blinded from clinical data and subsequently a final consensus was obtained under the supervision of a 20-year experienced pediatric neuroradiologist.
Patient clinical data sets were collected from online clinical records and stored in an anonymized database. Each patient performed every MRI during hospitalization, together with a complete clinical examination. For each MRI timepoint, we reviewed clinical data to infer disease-related disability as a global indicator of disease progression over time. Based on previous works , we defined patient disability as a variable with 3 grades (Table 1): autonomous unassisted ambulation (grade 0), autonomous ambulation with gait aid (grade 1), and absence of ambulation (patient on wheelchair) (grade 2). Furthermore, we collected epidemiological data from patient reports including age, disease duration, and genetic results. Disease onset was evaluated retrospectively based on KSS clinical diagnostic criteria, rather than genetic confirmation, because genetic testing was often delayed.
A chi-square test was performed to investigate the prediction value of patient disability with brain score and spinal cord involvement. For those variables having a significant chi-square test, a contingency coefficient, phi factor, and Cramer’s V were also computed. We set the significance threshold (p) to 0.05. Statistical analysis was performed with SPSS Statistics (v.21, IBM, N.Y., USA).