Target trial emulation and causal machine learning define individualized safety margins for energy exposure near the recurrent laryngeal nerve in thyroidectomy: multicentre cohort study.

BACKGROUND: The aims of this study were to evaluate whether cumulative thermal exposure from surgical energy devices near the recurrent laryngeal nerve (RLN) is associated with postoperative RLN injury (RLNI) and to derive clinically interpretable individualized safety margins. METHODS: This retrospective multicentre cohort study included adults who underwent thyroidectomy between 2018 and 2024. The thermal accumulation index (TAI) quantified cumulative activation-level energy exposure within a 2-mm nerve-dense zone. Target trial emulation, generalized propensity score adjustment, doubly robust causal estimation, and causal machine learning were used to estimate population-level and individualized effects. RESULTS: Among 7145 patients, after adjustment for prespecified baseline patient-, disease-, and surgical-domain covariates, as well as PCA-derived proxies for surgeon skill, tissue vulnerability, anatomical variation, and anaesthesia-related factors, a higher TAI was associated with an increased risk of RLNI, with an approximately linear dose-response curve over the observed range (odds ratio 3.01; 95% confidence interval 2.35 to 3.87). In the independent test cohort, the observed RLNI rate was 2.3%; the model-projected rate under a low-TAI strategy was 0.8%, corresponding to a projected relative reduction of 65.9%. CONCLUSION: TAI provides a candidate quantitative metric for cumulative thermal exposure near the RLN. The individualized thresholds generated in this study should be considered hypothesis-generating and require prospective validation before routine intraoperative use.