Forecasting stroke-like episodes and outcomes in mitochondrial disease

In this retrospective, multicentre, observational cohort study, we sought to determine the clinical, radiological, EEG, genetics and neuropathological characteristics of mitochondrial stroke-like episodes and to identify associated risk predictors. Between January 1998 and June 2018, we identified 111 patients with genetically determined mitochondrial disease who developed stroke-like episodes. Post-mortem cases of mitochondrial disease (n = 26) were identified from Newcastle Brain Tissue Resource. The primary outcome was to interrogate the clinico-radiopathological correlates and prognostic indicators of stroke-like episode in patients with mitochondrial encephalomyopathy, lactic acidosis and stroke-like episodes syndrome (MELAS). The secondary objective was to develop a multivariable prediction model to forecast stroke-like episode risk. The most common genetic cause of stroke-like episodes was the m.3243A>G variant in MT-TL1 (n = 66), followed by recessive pathogenic POLG variants (n = 22), and 11 other rarer pathogenic mitochondrial DNA variants (n = 23). The age of first stroke-like episode was available for 105 patients [mean (SD) age: 31.8 (16.1)]; a total of 35 patients (32%) presented with their first stroke-like episode >= 40 years of age. The median interval (interquartile range) between first and second stroke-like episodes was 1.33 (2.86) years; 43% of patients developed recurrent stroke-like episodes within 12 months. Clinico-radiological, electrophysiological and neuropathological findings of stroke-like episodes were consistent with the hallmarks of medically refractory epilepsy. Patients with POLG-related stroke-like episodes demonstrated more fulminant disease trajectories than cases of m.3243A>G and other mitochondrial DNA pathogenic variants, in terms of the frequency of refractory status epilepticus, rapidity of progression and overall mortality. In multivariate analysis, baseline factors of body mass index, age-adjusted blood m.3243A>G heteroplasmy, sensorineural hearing loss and serum lactate were significantly associated with risk of stroke-like episodes in patients with the m.3243A>G variant. These factors informed the development of a prediction model to assess the risk of developing stroke-like episodes that demonstrated good overall discrimination (area under the curve = 0.87, 95% CI 0.82-0.93; c-statistic = 0.89). Significant radiological and pathological features of neurodegeneration were more evident in patients harbouring pathogenic mtDNA variants compared with POLG: brain atrophy on cranial MRI (90% versus 44%, P < 0.001) and reduced mean brain weight (SD) [1044 g (148) versus 1304 g (142), P = 0.005]. Our findings highlight the often idiosyncratic clinical, radiological and EEG characteristics of mitochondrial stroke-like episodes. Early recognition of seizures and aggressive instigation of treatment may help circumvent or slow neuronal loss and abate increasing disease burden. The risk-prediction model for the m.3243A>G variant can help inform more tailored genetic counselling and prognostication in routine clinical practice.

Automated summary

This study aimed to determine the characteristics of mitochondrial stroke-like episodes and develop a risk-prediction model. The m.3243A>G variant was found to be the most common genetic cause of stroke-like episodes. The clinical, radiological, EEG, and neuropathological features of stroke-like episodes were consistent with medically refractory epilepsy. Patients with different genetic variants had different disease trajectories and prognoses. The risk-prediction model for the m.3243A>G variant showed good performance. Patients with pathogenic mitochondrial DNA variants were more likely to have radiological and pathological features of neurodegeneration compared to those with POLG-related episodes.