Cluster Analysis to Specify the MELAS Genotype and Phenotype Must Consider Genetic Background Peculiarities and Disease Progression

We read with interest the article by Alves et al,¹ on a retrospective study of the genotypic and phenotypic patterns of 35 patients with genetically confirmed mitochondrial encephalopathy, lactic acidosis, and strokelike episodes (MELAS) syndrome, using unsupervised hierarchic cluster analysis. Cluster analysis was performed using 53 discrete variables that yielded 2 distinct phenotypes, classic MELAS and atypical MELAS.¹ It was concluded that the recognition of these 2 MELAS presentations will allow clinical and research communities to better understand the clinical presentation and prognosis of MELAS and to identify specific therapeutic interventions.¹ The study is impressive, but some points require further discussion.

The first point is that the study had a retrospective design.¹ Retrospective designs have the disadvantage that data can be missing, the accuracy of the data cannot be easily checked, desired missing or new data can no longer be generated, and indications for certain investigation are often not comprehensible. We should know how much data of the cohort were missing and to what extent these missing data affected the results. How many patients were excluded due to missing data? Other disadvantages of the design are that it was monocentric and uncontrolled.

A second point is that the cluster analysis did not consider several factors that significantly influence the phenotype. These include the mitochondrial DNA (mtDNA) copy number, haplotype, mtDNA polymorphisms, and nuclear background. The cluster analysis also did not consider that MELAS is typically a progressive disease, meaning that various manifestations develop during the disease course; these may or may not be present at a given cross-sectional assessment. Therefore, cluster analysis may produce different results at different times, explaining why it should be repeated at different times. Another disadvantage of the cluster analysis is that only a small number of patients were included.

The third point is that the included patients were not diagnosed with MELAS according to the Hirano or Japanese criteria usually used for this diagnosis.^2,3 Therefore, we should know how many of the 35 patients met the Hirano or Japanese criteria.

The fourth point is that seizures as a classic phenotypic feature of MELAS and, in particular, a manifestation of strokelike episodes (SLEs) were not included in the cluster analysis. We should know how many of the 35 patients had seizures as a manifestation of SLEs or seizures without a SLE. It would also be interesting to know how many patients required regular antiseizure medication.

The fifth point is that the current medications were not included in the assessment. Because several drugs have mitochondrial-toxic potential, it is imperative to know which drugs the included patients were taking regularly. In particular, valproic acid, carbamazepine, barbituric acid, and zonisamide are potentially mitochondrial-toxic.⁴ Biguanides are also known to increase the serum and CSF lactate levels.

In summary, more important than identifying and delineating a syndrome is clarifying the genetic background and pathophysiology of a hereditary disease. Diagnostic and therapeutic management may depend more on the etiology and pathogenesis than on the phenotypic variability.

• © 2024 by American Journal of Neuroradiology