Intraparotid lymph node metastasis affects distant metastasis in parotid adenoid cystic carcinoma

To evaluate the relationship between factors of metastatic intraparotid lymph node (IPLN) and distant metastasis in parotid adenoid cystic carcinoma (ACC). Patients with surgically treated parotid ACC were retrospectively enrolled, and primary outcome variable was distant metastasis free survival (DMFS). The effect of factors of metastatic IPLN on DMFS was evaluated using Cox model. In total, 232 patients were included. Extranodal extension of IPLN and cervical lymph nodes did not impact the DMFS, and the 7th but not 8th AJCC N stage was associated with DMFS. Groups of 0 and 1 metastatic IPLN had comparable DMFS, but presence of 2+ positive IPLN was related to increased worse DMFS (p = 0.034, HR 2.09). A new N stage (0 vs 1–2 vs 3+) based on total positive lymph node number exhibited better C-index than traditional N stage. IPLN metastasis increased the risk of distant metastasis, and the impact was mainly determined by the number of metastatic IPLN. Our proposed N stage provided better DMFS prediction than the 8th AJCC N classification.


09). A new N stage (0 vs 1-2 vs 3+) based on total positive lymph node number exhibited better C-index than traditional N stage. IPLN metastasis increased the risk of distant metastasis, and the impact was mainly determined by the number of metastatic IPLN. Our proposed N stage provided better DMFS prediction than the 8th AJCC N classification.
Salivary gland carcinomas are relatively uncommon, and account for less than 3% of all head and neck cancers 1 , most of them occur in parotid gland. One of the most frequent pathologic types is adenoid cystic carcinoma (ACC), it is featured by distant metastasis (DM) 2 . Complete resection is the preferred method, but DM is likely to develop in 40% or more of the patients and also the main cause of death during follow-up 3 . To detect the potential predictors for DM carries essential significance to improve the oncologic outcome by filtering high risk patients.
At present, prior evidence has described that perineural invasion (PNI), lymphovascular invasion (LVI), cervical lymph node (LN) metastasis, and positive margin increase the risk of DM [4][5][6][7][8][9] , but the role of intraparotid lymph node (IPLN) is rarely discussed, to our best knowledge, only one research has reported the association between IPLN metastasis and DM in parotid cancer 10 , in this study, positive IPLNs were noted in 31.8% of the sample and provided nearly additional onefold possibility of DM compared with non-metastasis group in Cox model analysis. But there is still much unknown regarding how burden and extranodal extension (ENE) of metastatic IPLNs impact the DM risk.
Therefore, the goal of current study was to evaluate the relationship between factors of metastatic IPLN and DM in parotid ACC.

Patients and methods
Ethical consideration. This study was approved by China Medical University Institutional Research Committee, all methods were performed in accordance with the relevant guidelines and regulations, and written consent agreements for medical research were obtained from all patients before the initial treatment.
Study design. The investigators performed a retrospective study to address the issue. From January 2000 and December 2022, medical records of patients with surgically treated parotid ACC were reviewed. Inclusion criteria were presented as follows: the disease was primary; pathologic section was available for re-reviewing; follow-up data could be obtained; neck dissection or sentinel pathologic examination of at least four cervical LNs were conducted 11 . Patients with a history of other malignancy or DM at initial treatment were excluded. Information regarding demography, pathology, treatment and follow-up of included patients was extracted. www.nature.com/scientificreports/ Study variable. All pathologic sections were re-reviewed by at least two head and neck pathologists to confirm the correct diagnosis of ACC. IPLN referred to the LNs located within the parotid gland. Tumor stage was formulated based on the 8th AJCC system, neck classification were formulated based on the 7th and 8th AJCC system, pathologic grade was defined as low for tubular or cribriform type, in which a solid component accounted for less than 30%, and high for solid type 12 13 . Its time was calculated from the date of surgery to the date of DM detection or the last follow-up. For all patients after treatment completion, they were followed up every 3-6 months for the first two years, every 6-12 months for the next three years, and every 12-24 months thereafter. Statistic analysis. The Kaplan-Meier method was used to compare the DMFS rates in patients with different factors of IPLN, and the factors which were significant in univariate analyses were then evaluated in Cox proportional hazard regression analyses to determine the independent predictors. Three Cox models were constructed and compared using C-index, a higher C-index meant better prognosis prediction, number of positive IPLNs was analyzed in model 1, the 7th AJCC N stage and number of positive IPLNs were analyzed in model 2, total number of positive LNs was analyzed in model 3. All analyses were manipulated using R 3.4.3, and a p less than 0.05 was considered to be significant.

Ethics approval and consent to participate. This study was approved by China Medical University
Institutional Research Committee, and written consent agreements for medical research were obtained from all patients before the initial treatment.

Association between IPLN metastasis and neck stage. In
Predictors for DM. During our follow-up with mean time of 4.2 ± 2.5 years, distant metastasis occurred in 125 (53.9%) patients, and the mean time of DM development was 3.1 ± 1.5 years, the overall 5-year DMFS rate was 46% (95% CI 38-54%). Lung was the most common metastasized site and developed in 100 patients, Table 1. Association between intraparotid lymph node metastasis and neck stage. www.nature.com/scientificreports/ of whom 74 patients had only lung distant and 26 patients had other metastasis sites simultaneously (bone: 14 cases; liver: 12 cases; brain: 5 cases; renicapsule: 3 cases; skeletal muscle: 1 case). In the rest 25 patients, bone metastasis occurred in 12 patients, liver metastasis occurred in 10 patients, and brain metastasis occurred in 6 patients.
Model 1 included tumor stage, pathologic grade, PNI, positive margin, and number of metastatic IPLNs. Zero and one positive IPLN groups had comparable DM risk, but presence of 2 or more metastatic IPLNs was related to about twofold risk (95% CI 1.18-5.34) of DM. Other independent factors included T4 stage, high grade, and positive margin ( Table 3). The C-index was 0.69 (95% CI 0.65-0.73).
Model 2 included tumor stage, the 7th neck stage, pathologic grade, PNI, positive margin, and number of metastatic IPLNs. Compared with no IPLN metastasis group, presence of 1 positive LN did not impact the DM, but presence of 2 or more metastatic IPLNs was associated with nearly threefold risk (95% CI 1.26-6.48) of DM. N1 and N0 stages had comparable possibility of DM, but N2/3 stage predicted significantly higher risk of DM (HR: 3.21; 95% CI 1.76-8.41). Other independent factors included T3/4 stage, high grade, and positive margin ( Table 3). The C-index was 0.74 (95% CI 0.72-0.76).

Discussion
Our most important finding was that IPLN metastasis increased the risk of DM, and the impact was mainly determined by the number rather than ENE of metastatic IPLNs. The 7th but not 8th AJCC N stage was related to DM development, prognostic model based on total number of metastatic LNs provided better survival stratification than those according to number of metastatic IPLNs and/or the AJCC N stage. Metastatic IPLN was an important prognostic factor in parotid cancer, it was related to decreased survival 14,15 , but DMFS was analyzed as an outcome variable only in a few studies 10,16 , the risk of DM was significantly increased if there was presence of metastatic IPLN, and the HR ranged from 2 to 4, however, all these researches evaluated the IPLN status as a binary variable (Yes vs No), it remained unclear regarding the effect of positive IPLN burden on prognosis. Feng et al. 17 might be the only one to evaluate the association between different positive LN number of parotid and survival in 337 patients, compared with no IPLN metastasis, 1 or 2 positive LNs linked with nearly sixfold risk of local recurrence, and 3 or more LNs had a HR of 21, but the authors failed to report the impact on DM. Our study firstly noted groups of 0 and 1 metastatic IPLN had similar DMFS, and presence of 2+ positive LNs predicted poorer control of DM. The finding posed meaningful clinical significance, because for most solid cancers, worse oncologic outcome would be expected even if there was only one metastatic LN, and aggressive treatments were likely to be given to these patients to improve the prognosis, such as www.nature.com/scientificreports/ adjuvant radiotherapy would be suggested for parotid ACC with no other adverse pathologic features but only one positive IPLN, our study may alter the senseless practice. The underlying mechanism could be explained by the relatively low LN metastasis frequency and slow-growing of parotid ACC. Considering the significance of IPLN metastasis, a suggestion of the inclusion of IPLN status in LN stage classification was proposed 18 , but this was not realized in the 8th AJCC N classification which was drafted based on head and neck squamous cell carcinoma 19 , prior literature described the official N stage could not well stratify the survival in salivary gland carcinoma 20 , our study also confirmed that the 8th AJCC N stage was not related to DM occurrence. It might be contributed by the distinct difference of biologic behaviour between the two kinds of disease, and the common death cause in parotid cancer was DM but not locoregional recurrence. Another factor that could not be ignored was ENE, which was usually an indicator for adjuvant chemotherapy to oppose the high possibility of DM in head and neck cancer, but it was false in parotid ACC based on our results. In a study including 114 patients with pN + salivary gland carcinoma 21 , ENE developed in 51% of the cases, and was related to PNI, LVI, advanced N stage, and higher number of positive LNs, but had no association with demography, tumor stage origin, and histology grade. After adjusting the number of positive cervical LNs, ENE did not impact the survival. In another similar study 22 , ENE occurred in 27 (40.9%) patients, and ENE group had comparable locoregional-free survival, overall survival, and DMFS with those without ENE. The conclusion was also confirmed by Lombardi et al. 23 . These findings elucidated that ENE of cervical LN tended to demonstrate limited impact on prognosis, but was associated with some adverse pathologic features which drove the prognosis actually. The interesting discovery was also appropriate in ENE of IPLN, its presence did not add any supernumerary DMFS decrease, and it could explained the fact that the 7th but not 8th AJCC N stage was related to the DMFS, and the underlying mechanism for the firstly reported finding might be explained by the small anatomic size of IPLN, even a minimal lesion could easily break through the capsule.
An alternative LN stage was needed to better stratify the survival of parotid cancer. Aro et al. 20 introduced a LN stage based on the metastatic LN number (0 vs. 1-2 vs. 3-21 vs. 22 +) after analyzing the outcomes in 4520 patients undergoing neck dissection for salivary gland carcinoma, it provided better survival prediction than the 8th AJCC N stage. Another three-category LN stage according to the number of positive LNs and ENE was also superior to the 8th AJCC N stage in prognostic calculation 24 . However, the impact of IPLN on survival was neglected in the two studies, and the variable was neither incorporated into the proposed N stages nor analyzed in a regression model. Very few authors had evaluated the IPLN and neck stage as one variable. In a study consisting of 307 patients treated for salivary gland carcinoma 23 , owing to the failure of the 8th AJCC classification in overall survival stratification, the authors described two new LN systems according to the number of positive LNs (0 vs 1-3 vs 4+) and/or their maximum diameter (< 20 mm vs 20+ mm) showed better accuracy in survival prediction. Boon et al. 25 assessed the outcomes of 177 salivary duct carcinoma patients and noted that the absolute number of positive LNs (0 vs 1-2 vs 3-15 vs 16+), rather than the traditional cervical stage, was the only www.nature.com/scientificreports/ significant prognostic factor for overall survival in the multivariate analysis. It remained unknown whether such classification could apply for parotid ACC which had apparently different features with other parotid cancers.
In current study, we formulated a three-category LN stage with combination of metastatic parotid and cervical LNs, the system had the highest C-index among the three models in predicting DMFS, it was simple and suitable for clinical use effectively 26 . But it was related to increased demands for LN detection, and detection of a small IPLN was usually time-consuming and labor-intensive, and required cooperation of surgeon and pathologists, sometimes the entire parotid gland should be microscopically examined for accurate diagnosis. Limitation in current study must be acknowledged, first, this was a retrospective study, it had inherent bias; second, only ACC was analyzed, it was not clear whether the finding could be realized in other histologic types; third, this was a single institution study, before clinical application, further validation was required.

Conclusion
In summary, IPLN metastasis increased the risk of DM, and the impact was mainly determined by the number rather than ENE of metastatic IPLNs. Our proposed N stage provided better DMFS prediction than the 8th AJCC N classification.

Data availability
All data generated or analyzed during this study are included in this published article. And the primary data could be achieved from the corresponding author.