$$\rightleftharpoonup{xx}$$
$$\longleftharp{xx}$$,
$$\longrightharp{xx}$$,
Study design and overall schematic
The excluded participants did not meet predefined eligibility criteria, had incomplete baseline imaging data, or lacked standardized longitudinal follow-up. Among the 187 included participants, complete 12-month outcome classification was available for all cases included in the primary analysis, although isolated intermittent missing observations at individual follow-up visits were accommodated within the mixed-effects modeling framework (Rysunek 1). Using the prespecified 12-month response definition based on VRR, 146 nodules were classified as responders (VRR ≥ 50%) and 41 as non-responders (VRR < 50%) (Rysunek 1). This design enabled paired within-nodule comparisons (baseline as the internal control) and standardized longitudinal assessment across predefined follow-up time points.
Baseline participant and nodule characteristics
Baseline clinical characteristics are summarized in Table 1. Non-responders were older than responders (49.63 ± 8.01 vs 45.94 ± 7.88 years, P = 0.009) and had slightly higher baseline symptom and cosmetic burden (pressure symptom score 3.43 ± 0.57 vs 3.19 ± 0.63, P = 0.029; cosmetic concern score 3.01 ± 0.47 vs 2.79 ± 0.49, P = 0.011). Thyroid function was largely similar between groups, although TSH was modestly higher among non-responders (1.86 ± 0.31 mIU/L vs 1.74 ± 0.33 mIU/L, P = 0.039) (Table 1). These baseline differences indicate a less favorable clinical profile among non-responders prior to intervention. Baseline ultrasound showed significantly larger nodules in non-responders, including greater maximum diameter (33.87 ± 4.03 mm vs 29.31 ± 3.94 mm, P < 0.001) and baseline volume (12.04 ± 2.41 mL vs 8.33 ± 1.90 mL, P < 0.001) (Table 1). Nodule composition also differed, with a shift toward higher cystic component categories in non-responders (P < 0.001) (Table 1). These baseline ultrasound characteristics provide an anatomic context for subsequent volumetric and biomechanical response patterns. No participant demonstrated diffuse thyroid parenchymal abnormalities severe enough to meet the predefined exclusion criteria for diffuse thyroid disease.
Elastography measurements were obtained under standardized conditions with prespecified region-of-interest rules and acceptance criteria to minimize motion and compression artifacts. The use of baseline within-nodule measurements as internal references may help reduce the influence of interindividual variability and support the interpretation of longitudinal stiffness changes. Procedural characteristics are shown in Table 1. Consistent with larger baseline nodule burden, non-responders had longer procedure duration (21.13 ± 3.01 min vs 18.42 ± 2.79 min, P < 0.001) and higher delivered energy (13.18 ± 2.63 kJ vs 11.04 ± 1.98 kJ, P < 0.001) (Table 1). These procedural differences were expected, given baseline size and composition differences and were accounted for in adjusted analyses of response prediction.
Longitudinal elastography trajectories after RFA
Immediately after RFA, both elastography modalities demonstrated a marked acute stiffness increase (Table 2; Rysunek 2), consistent with the hypothesized immediate biomechanical effect of thermal coagulation. Shear-wave elastography (SWE) increased from 31.5 ± 5.1 to 41.7 ± 5.6 kPa in responders and from 35.1 ± 5.3 to 43.4 ± 5.8 kPa in non-responders (Table 2; Rysunek 2). Strain ratio increased concordantly (responders 2.41 ± 0.37 to 3.14 ± 0.41; non-responders 2.77 ± 0.40 to 3.35 ± 0.42) (Table 2; Rysunek 2). In contrast, immediate post-ablation volume remained close to baseline in both groups (Table 2), supporting the study hypothesis that elastography provides an earlier, early imaging indicator of treatment-related tissue change than short-interval volumetry.
Longitudinal volumetric outcomes and VRR
Over follow-up, elastography demonstrated progressive remodeling after the immediate post-ablation rise (Table 2; Rysunek 2). In responders, SWE fell below baseline by 3 months and continued to decline through 12 months (31.5 ± 5.1 kPa at baseline to 21.6 ± 4.7 kPa at 12 months), whereas non-responders showed an attenuated reduction and remained substantially stiffer at 12 months (31.2 ± 5.2 kPa) (Table 2; Rysunek 2). Strain ratio followed the same pattern, decreasing to 1.87 ± 0.35 at 12 months in responders versus 2.55 ± 0.41 in non-responders (Table 2; Rysunek 2).
Volumetric outcomes evolved in parallel with biomechanical changes (Table 2; Rysunek 3). Nodule volume decreased steadily in both groups, with persistent separation across time points; at 12 months, responders reached 2.20 ± 0.45 mL compared with 7.35 ± 1.80 mL in non-responders (Table 2; Rysunek 3). VRR also separated early and remained distinct: at 1 month, VRR was 31.5 ± 6.3% in responders versus 16.1 ± 3.8% in non-responders, and at 12 months VRR was 73.6 ± 6.9% versus 38.5 ± 8.7%, respectively (Table 2; Rysunek 3). Together, these aligned elastography and VRR trajectories support the hypothesis that post-RFA biomechanical remodeling tracks and differentiates durable volumetric response.
Early elastography change as an independent predictor of 12-month volumetric response
Mixed-effects models corroborated the descriptive trajectories and quantified group differences over time (Table 3). For SWE, time effects indicated a significant immediate increase relative to baseline (estimate +10.22 kPa, P < 0.001) followed by progressively lower values at later visits (e.g., 12 months vs baseline estimate -9.02 kPa, P < 0.001). Importantly, time-by-group interactions showed divergence emerging at later follow-up, with non-responders exhibiting higher SWE relative to responders from 3 months onward (non-responder × 3 months estimate +4.20 kPa, P < 0.001; non-responder × 12 months estimate +9.41 kPa, P < 0.001) (Table 3). Strain ratio models demonstrated the same pattern, with significant late divergence (non-responder × 12 months estimate +0.68