Materials and methods

Patient and healthy control recruitment criteria
Nine adult patients referred to the NHS Specialised Services-funded Mitochondrial Diagnostic Centre in London for investigation of CPEO and found to harbour a single, sporadically occurring, mtDNA deletion in skeletal muscle, and nine age- and gender-matched healthy controls were recruited. The mean age of the patients was 33 years (range: 19–52 years) with a mean age at symptom onset of 16 years (range: 12–32 years). Four patients with a clinical diagnosis of Kearns-Sayre syndrome (KSS) had a mean age at onset of 12 years (range: 6–17 years). The remaining patients with CPEO had a mean age of onset of 20 years (range: 14–32 years). The mean age of the controls was 33 years (range: 26–43 years).
Institutional Research Ethics Committee approval for the study, and accordingly individual subjects’ consent, was obtained.

Clinical assessment of ocular movements
EOM function was quantified in terms of ROEM by measuring ocular excursions (ductions) for each eye on the Goldmann perimeter using, as in a previous CPEO study [8], the modified uniocular field of fixation technique [9]. This isolates the action of each individual muscle by recording excursions only along the axes corresponding to that muscle’s primary field of action. These are plotted on standard Goldmann charts with pre-printed testing axes. For the right eye the axes are: 0° (lateral rectus, LR); 67° (superior rectus, SR); 141° (inferior oblique, IO); 180° (medial rectus, MR); 216° (superior oblique, SO); and 293° (inferior rectus, IR). The left eye uses a chart which is the mirror image of the right eye chart. Testing was performed uniocularly with chin and head fixation and excursions repeated to ensure measurement consistency within 5°.

Magnetic resonance imaging protocol
Imaging was performed at 3 T (TIM Trio, Siemens, Erlangen, Germany) with a 32-channel receiver head coil. Subjects were examined supine and headfirst and asked to keep their eyes closed for the duration of each scan. Coronal images were acquired with a field of view (FOV) of 162 × 200 mm. Turbo-spin echo (TSE) T1w images were acquired with time to repetition (TR)/time to echo (TE) = 575/7.6 ms, 20 × 2-mm slices, slice gap 1 mm, matrix 312 × 384. STIR images were acquired with TR/TE/time to inversion = 6760/8/220 ms, matrix 312 × 384 and 20 × 2-mm slices, slice gap 1 mm. T2 was measured with a multiple spin-echo sequence (TR = 4060 ms, 22 spin-echoes with TEs 8.6–189.2 ms and an echo spacing of 8.6 ms, matrix 208 × 256, 12 × 3-mm slices, 0.3-mm gap, 180° refocusing pulses). Total imaging time was less than 20 min.

Qualitative magnetic resonance imaging evaluation
All MRI analysis was performed by observers blinded to the subject clinical details, including patient versus control status. The coronal T1w and STIR orbital images were systematically reviewed by two experienced neuroradiologists (I.D. and T.Y.). The individual characteristics of all 12 EOMs were reviewed sequentially in all subjects and the consensus opinion was recorded. A scoring system was applied to capture the following qualitative features: degree of atrophy (normal = 0, mild = 1, marked = 2); STIR hyperintensity (normal = 0, hyperintense = 1); and presence or absence of T1w hyperintensity foci with anatomically coincident STIR hypointensity (presumed to represent fatty infiltration), taking account of the intra-EOM pattern: the intrinsic foci were observed to either be of approximately equal diameter in all dimensions in the coronal plane, termed 'dots', or elongated in one dimension (invariably in-line with the longest dimension of the extra-ocular muscle in the coronal section), termed 'streaks’. Where both dots and streaks were observed in different sections through a single EOM, the convention adopted was to record the muscle as demonstrating an intrinsic 'streak' signal. The qualitative scores were tabulated for both healthy controls and CPEO patients.

Measurement of cross-sectional area
Cross-sectional areas of the five EOMs bilaterally (excluding IO) were obtained from the coronal T1w images by outlining regions of interest (ROI) around the outer border of each muscle, one slice behind the posterior border of the globe, using the FSLView software (FMRIB, Oxford, UK), and the cross-sectional area of each was recorded. All measurements were performed by a single observer (R.D.S.P) and reviewed by a second observer (I.D.) for quality control purposes.

Quantitative magnetic resonance analysis
A single observer (J.M.) defined additional ROIs on the TE = 8.6 ms multi-echo image of the first coronal slice posterior to the globe, for the four rectus muscles bilaterally using the ITK-SNAP software [10]. Only voxels wholly within the muscle border were included to avoid partial volume errors arising from pixels containing both muscle and orbital fat. The small size of the superior oblique EOM prevented reliable ROI placement and this EOM was therefore omitted from the analysis. The mean ROI signal intensities were calculated for each TE image and fitted to a mono-exponential decay function to obtain the intercept, T2 decay constant and a DC offset using the Levenberg-Marquardt algorithm implemented in Wolfram Mathematica (Champaign, IL, USA).

Statistical methods
As the different analyses included different muscle subgroups (all 12 muscles for ROEM analyses, ten muscles for a cross-sectional area, eight muscles for T2 measurement) when overall means were calculated only the common set of muscles were included, left and right: SR, IR, LR and MR. Mean values in patient and control groups, reported with their respective standard deviations, were compared using the two-tailed Student’s t-test. The proportion of muscles in each group with T1w hyperintensity was compared with Pearson’s χ2 test. Correlations between MRI measures were analysed using the Spearman rank correlation (rho). All statistical analyses were performed using IBM SPSS Version 20.