Abstract
Clinical recognition of primary mitochondrial disorders (PMD) is difficult due to the clinical and genetic heterogeneity. Whereas lactate has low sensitivity and specificity, in structured clinical studies growth differentiation factor 15 (GDF15) has shown promise with elevations in mitochondrial DNA (mtDNA)-related PMD, but its specificity has been questioned. In a tertiary care hospital-wide study, medical records were retrospectively reviewed from 418 cases where GDF15 levels were obtained by clinicians. Patients were classified into patients with PMD due to mtDNA-related defects (mtDNA maintenance, mtDNA deletions, and mtDNA-encoded tRNA variants), PMD due to structural defects or other nuclear causes, and in non-mitochondrial disease. Patients with liver disease or systemic critical illness were excluded. GDF15 was assayed in a clinical laboratory with a cutoff of 750 ng/L. There were 38 mtDNA-related PMD (GDF15 >750 pg/mL in 76%), 35 other nuclear DNA-encoded PMD or structural subunits (31% elevated GDF15), 309 non-mitochondrial disorders (13% elevated GDF15). Based on the highest Youden J-index, the optimal cut-off value to identify these target mtDNA-related disorders was 815 pg/mL, with sensitivity 76%, specificity 88%, positive predictive value of 41% and negative predictive value of 97%. At this optimized cutoff level, mtDNA-encoded PMD patients had elevated GDF15 in 76%, nuclear DNA-encoded PMD in 26%, and non-mitochondrial disorders in 11% of patients. Thus, in a real-life clinical setting, after excluding abnormal liver function and critical illness, GDF15 had good clinical utility increasing the odds at predicting mtDNA-related primary mitochondrial disorders 14-fold, but not for structural or other nuclear-encoded primary mitochondrial disorders.