Endpoint Retention in Public POLG Deoxycytidine-Deoxythymidine Trial Reports

**Background:** POLG-related disease is caused by pathogenic variants in the gene encoding DNA polymerase gamma, but a clinical endpoint story cannot be inferred merely because several public endpoints move in a favorable direction. Public deoxycytidine/deoxythymidine reports now include a 2024 six-month interim POLG trial report and a 2026 longer-term report from the same trial family, creating a specific endpoint-retention problem rather than a generic mitochondrial-treatment question. **Objective:** The selected question was whether longitudinal Newcastle Mitochondrial Disease Scale (NMDS), growth differentiation factor 15 (GDF-15) and quality-of-life endpoints in public POLG deoxycytidine/deoxythymidine reports support sustained endpoint concordance, or whether deaths, withdrawals, missing biomarker windows and single-arm cohort overlap make concordance non-identifiable. **Methods:** Complete public primary texts and indexed primary records were inspected for two POLG dC/dT trial-family reports, two POLG-relevant preclinical nucleoside sources, one non-POLG/non-TK2 dC/dT clinical report and one 2026 POLG biomarker cohort. A source-comparison pilot coded endpoint class, follow-up window, denominator or event caveat and claim boundary. No patient-level data, private platform inputs, pooled effect size, treatment ranking, survival model or clinical recommendation was generated. **Results:** Six public endpoint rows from two overlapping POLG trial-family reports were coded across NMDS, GDF-15, EEG and quality-of-life classes; five rows had favorable reported directions and four rows were retention- or measurement-limited. The 2024 first-ten report recorded NMDS and GDF-15 improvement over six months, while the 2026 expanded report recorded 25 starters, five deaths, five withdrawals, improved NMDS through reported windows, early GDF-15 improvement and quality-of-life improvement at selected timepoints. The same evidence also preserved open-label single-arm design, cohort overlap, changing timepoint support, missing GDF-15 windows and upper-limit biomarker constraints. Preclinical nucleoside studies supported biological plausibility and endpoint-transfer caution, not clinical rescue. **Conclusions:** Public POLG dC/dT reports support endpoint-specific favorable signals under important retention conditions, not a demonstrated same-participant concordant clinical rescue claim. The analysis is non-causal, non-pooled and hypothesis-generating: future POLG dC/dT interpretation should require matched participant/time windows, event and withdrawal accounting, biomarker censoring rules, genotype/onset strata and comparator or natural-history anchoring before endpoint concordance is treated as identifiable. AI tools assisted literature synthesis, drafting, data visualization, formatting and QA preparation. Gemini independently reviewed the exact rendered PDF. Andrew Friedl remains responsible for the content and is the sole author.