Evidence-domain translation in the Lyme disease controversy over persistent post-treatment symptoms.

The debate between post-treatment Lyme disease syndrome (PTLDS) and chronic Lyme disease (CLD) reflects different views about the causes and treatment of persistent symptoms attributed to Lyme disease, including symptoms that continue after recommended antibiotic treatment. This study examines how competing positions in the controversy draw on different scientific evidence domains, and how findings from those domains are extended into broader clinical and mechanistic claims. We analysed a 2000-2024 literature corpus using a prompt-optimised three-model large language model ensemble to classify abstracts by stance, theme and model-derived study design. We then conducted targeted full-text audits of selected retreatment and treatment-duration trials, and preclinical persistence studies. Across the higher-volume evidence domains, model-derived claim orientations differed by study-design tier. Observational studies and commentaries showed PTLDS-oriented distributions, whereas case reports or series, animal models, and in vitro studies showed CLD-oriented distributions. The smaller RCT and guideline tiers were also PTLDS-oriented, but the small numbers and wide confidence intervals made this direction uncertain. The clinical trial audit showed that some trials reported short-term or symptom-specific improvements, while the interpretation of these findings depended on their durability, endpoint consistency, eligibility criteria, treatment burden, and safety. The citation-conditioned preclinical audit identified findings that support several candidate mechanisms and generate hypotheses for further study, but these findings did not by themselves establish viable infection as the cause of persistent human symptoms or demonstrate the efficacy of prolonged antimicrobial treatment. The findings show a recurring distinction between mechanism-level evidence and the evidence needed to establish patient-level causation or durable clinical benefit. They map how these different forms of evidence are distributed and translated across the PTLDS-CLD literature.