Discussion

Although thyroidectomy is a well-established procedure, it carries inherent risks of postoperative complications. Postoperative hypocalcemia is the most common one [21]. Managing such complications requires accurate intraoperative identification of anatomical structures and prompt postoperative interventions, such as calcium and vitamin D supplementation [21]. Furthermore, while recent literature demonstrates that total thyroidectomy (TT) for differentiated thyroid carcinoma does not significantly increase the risk of early complications like hypoparathyroidism compared to subtotal thyroidectomy (STT), the overall incidence remains a substantial clinical challenge regardless of the surgical extent [22].

NIRAF has emerged as a transformative tool in endocrine surgery, significantly enhancing the intraoperative identification and preservation of PGs compared to traditional subjective visual inspection, and reducing the incidence of postoperative hypoparathyroidism [9]. For instance, a multicenter RCT demonstrated that NIRAF significantly increased the identification rate (47.1% vs. 19.2%, P < 0.001) and markedly reduced the risk of postoperative hypocalcemia (9.1% vs. 21.7%, P = 0.007) [5]. Our findings strongly confirm this clinical value. The NIRAF group demonstrated significantly higher postoperative PTH levels (P = 0.002) and serum calcium levels (P = 0.031) than those in the non-NIRAF group. This aligns with recent high-quality evidence, including comprehensive meta-analyses and multicenter randomized clinical trials (RCTs) [5], [7], [16]. Ultimately, NIRAF serves as a highly precise navigation tool that provides real-time guidance, translating into superior parathyroid preservation and improved biochemical stability for patients.

Despite its proven clinical benefits and high overall accuracy, the diagnostic precision of NIRAF is unsatisfactory, even in high-volume centers. A clinical trial by Kiernan et al. reported a false-positive rate of 11.1% due to overlapping intrinsic fluorescence from adjacent structures [11]. Similarly, another prospective study in 2023 reported a false-negative rate of 19.6% using NIRAF [23]. In our NIRAF group, we observed a false-negative rate of 20% and a false-positive rate of 16%, indicating that anatomical and optical interferences can still lead to inadvertent parathyroidectomy. Interestingly, our comparison revealed that the rate of inadvertent parathyroidectomy in the NIRAF group was not significantly different from that in the Non-NIRAF group (20% vs. 32%, P > 0.05). This suggests that false negatives do not inevitably contribute to inadvertent excision. To definitively minimize the inadvertent parathyroidectomy rate, Kuo TC et al. pointed that rather than relying exclusively on NIRAF, combining autofluorescence with adjunctive methods, such as PTH test strips or staining techniques, may be necessary to overcome these diagnostic limitations [24]. Therefore, to optimize the clinical application of this technology, a detailed investigation into the specific biological and technical causes of these inaccuracies is warranted.

The identification procedure for NIRAF is affected by false-negative signals. In our multivariate analysis, a higher BMI was associated with an increased incidence of inadvertent parathyroidectomy (P = 0.069). This assumption has several significant clinical implications. Excessive cervical fat accumulation substantially increases the risk of false-negative outcomes. Thick adipose tissue restricts the penetration depth of autofluorescence signals by altering light scattering [23]. Previous studies have reported that the maximum detectable depth of NIRAF reaches up to 3.05 mm [23]. Consequently, the inherently weak fluorescence emitted by true PGs enveloped in deep adipose tissue is easily obscured or attenuated, rendering them invisible on the display and ultimately increasing the risk of inadvertent excision during surgery [25], [26].

False-positive signals originating from surrounding nontarget tissues also impair diagnostic specificity and cause visual confusion. Our study documented four representative false-positive tissues, including thyroid nodules, adipose tissue, lymph nodes, and thymic tissue (Fig. 1). These interferences stem from the overlapping intrinsic optical and biological properties of these tissues. First, hyperplastic thyroid nodules exhibit abnormally enhanced autofluorescence, known as the “white light effect,” which is primarily associated with their internal colloid content and local vascular proliferation [18], [27], [28], [29]. Second, specific cervical adipose tissues contain intrinsic fluorophores that generate bright visual artifacts within the surgical field [30]. Third, reactive or metastatic lymph nodes present an endogenous optical signature similar to that of PGs, driven by their increased metabolic activity and nonspecific retention of fluorophores [18], [27], [31]. Finally, thymic tissue can exhibit confounding false-positive autofluorescence. Anatomically and embryologically, the thymus shares a common origin with inferior PGs, which may contribute to highly similar intrinsic fluorophore compositions that mislead visual assessment [32].

The hardware configuration, specifically probe-based versus camera-based systems, has been observed to have varying effects on the occurrence of both false positives and false negatives during parathyroid identification. Regarding probe-based systems, existing data indicate high sensitivity but limited impact on functional preservation. A 2025 multicenter RCT by Cousart et al. demonstrated improved PG identification (3.3 vs 2.8, P < 0.001) without significant difference in postoperative hypoparathyroidism rates [7]. Furthermore, this probe system significantly reduced the reliance on frozen section biopsies (4.0% vs. 11.2%, P = 0.01), thereby reducing false-positive excisions [7]. In contrast, camera-based imaging systems, which were utilized in our study, exhibited slightly different clinical outcomes. Benmiloud et al. reported that global spatial visualization provided by the camera-based system effectively mitigated inadvertent parathyroidectomy. Their multicenter RCT reported that this system significantly reduced inadvertent parathyroidectomy (2.5% vs. 11.7%, P = 0.006) [5]. A 2025 meta-analysis showed that the estimated overall accuracy of image-based methods is 0.96 (95% CI, 0.87–0.99), while the estimated overall accuracy of probe-based methods is 0.93 (95% CI, 0.92–0.94) [33]. The statistical significance was 0.36, indicating that there is insufficient evidence to indicate a significant difference in overall accuracy, sensitivity, and specificity between these two methods [33]. However, this meta-analysis also indicated that camera-based device offers distinct advantages in contactless global visualization, potentially favoring functional preservation, though direct comparative high-level evidence remains limited [33]. Some researches indicated that utilizing a 760–770 nm laser wavelength may be more conducive to improving diagnostic accuracy in camera-based systems [34], [35]. Moreover, the energy platforms utilized, such as the harmonic scalpel (HS) employed in our study, plays a crucial role in optimizing NIRAF accuracy. While both HS and LigaSure (LS) exhibit identical safety profiles regarding major bleeding, HS is significantly more effective at reducing minor bleeding, particularly in thyroid carcinoma (P = 0.02), according to the previous study [36]. This superior hemostatic performance minimizes blood exudation, thereby preventing false negatives caused by obscuration of the surgical field; meanwhile, compared to LS, HS features a finer tip that better limits thermal damage to surrounding tissue, consequently reducing nonspecific autofluorescence and false-positive visual interference. Future prospective comparative studies are required to definitively determine the optimal technological approach.

This study has several limitations. First, although we attempted to minimize bias through strict 1:1 matching of cases, this single-center design remains subject to potential selection bias. Second, the relatively small sample size (n = 50) limits the statistical significance of the multivariate analyses. It is necessary to verify secondary trends in a larger cohort. Finally, long-term complications were not fully assessed because of the short-term follow-up. To validate our conclusions, future prospective multicenter studies are essential to validate these findings.

In summary, this study confirmed that the NIRAF is a highly sensitive navigation tool that improves PG preservation and postoperative outcomes. However, its diagnostic accuracy is challenged by false positives from heterogeneous tissues and false negatives driven by anatomical barriers such as excessive adipose tissue. To address these challenges and enhance overall precision, future strategies could consider integrating multimodal platforms and optimizing surgical workflows through standardized protocols, thereby further elevating the safety of endoscopic thyroid surgery [31], [37], [38].