Patients’ characteristics and outcomes

The characteristics of the patients are summarized in Supplementary Table S1. This study included a total of 369 patients with a median age of 47 years (range, 19–70 years). The median follow-up time was 51 months (range, 4–105 months). At the end of follow-up, 33 patients (8.9%) had died, and 48 patients (13%) had experienced disease relapse in the form of local recurrence (34 patients, 9.2%), regional failure (20 patients, 5.4%), and distant metastasis (47 patients, 12.7%). The 3-year LRFS, RRFS, DMFS, PFS, and OS rates were 93.2%, 96.2%, 88.9%, 81.3%, and 94.6%, respectively.

SUVmax and clinical stage

The SUVmax-t and SUVmax-n increased with the T and N stage, respectively (Fig. 1). When the values were distributed according to the disease stage, we found that the SUVmax-t data were more concentrated than the SUVmax-n data. The median SUVmax-t for the T1, T2, T3, and T4 stages were 6.2, 7.7, 9.4, and 11.2, respectively. The median SUVmax-n for the N0, N1, N2, and N3 stages were 1.6, 5.6, 7.1, and 9.1, respectively. The SUVmax-t and SUVmax-n significantly differed between different T stages and between different N stages, respectively (P < 0.001, Fig. 1). However, the SUVmax-t was not significantly correlated with the SUVmax-n (r = 0.13, P < 0.001, Supplementary Figure S1).

Fig. 1
figure 1

Violin plots. A Distribution of the SUVmax-t under T stage; B Distribution of the SUVmax-n under N stage

Optimal cutoff points for SUVmax-t and SUVmax-n

The mean SUVmax-t of the primary tumor and the mean SUVmax-n of cervical lymph node metastases were 9.2 ± 5.2 (range, 1.4–49.6) and 6.8 ± 5.8 (range, 0–36.6), respectively. The optimal cutoff SUVmax-t for predicting local recurrence was 7.5 (area under the curve [AUC] = 0.627, P = 0.015; Fig. 2). The optimal cutoff SUVmax-n for predicting regional recurrence was 6.9 (AUC = 0.757, P < 0.001). With an SUVmax-t threshold of 7.5, we could correctly identify approximately 82% of the patients with local recurrence. Moreover, at this cutoff, we could also identify approximately 44% of the patients with no risk of local recurrence (P < 0.001, Fig. 2A). With an SUVmax-n threshold of 6.9, we could correctly identify approximately 85% of the patients with regional recurrence and approximately 59% of the patients with no risk of regional recurrence (P < 0.001, Fig. 2B).

Fig. 2
figure 2

Probability density functions. A the SUVmax-t for predicting local recurrence; B the SUVmax-n for predicting regional recurrence

To control for potential confounders, we built a multivariable Cox proportional hazards model including all relevant variables. The results showed that SUVmax-t was an independent predictor of LRFS (HR = 3.741, 95% CI: 1.489–9.396, P = 0.005), while SUVmax-n was a risk factor for RRFS (HR = 3.238, 95% CI: 1.103–9.505, P = 0.033; Supplementary Table S2).

SUVmax and prognosis

Considering the above results, we stratified patients into 4 groups: (a) LL group, low SUVmax-n (≤6.9) and low SUVmax-t (≤7.5), (b) LH group, low SUVmax-n (≤6.9) and high SUVmax-t (>7.5), (c) HL group, high SUVmax-n (>6.9) and low SUVmax-t (≤7.5), and (d) HH group, high SUVmax-n (>6.9) and high SUVmax-t (>7.5). To better explain the results, we demonstrated several representative 18F-FDG PET images for each subgroup (Supplementary Figure S2, S3, S4 and S5).

The cumulative survival curves for the LH, HL, and HH groups were very close, but these curves were clearly separated from the survival curve for the LL group (P = 0.044; Fig. 3). The OS rate of the LL group significantly differed from those of the other 3 groups. Therefore, we reorganized all patients into 2 groups: a high-risk group consisting of the patients in the LH, HL, and HH groups, and a low-risk group consisting of the LL group patients. We found that the high-risk group had worse RRFS, LRFS, PFS, and OS rates than the low-risk group (all P < 0.05; Supplementary Figure S6).

Fig. 3
figure 3

Kaplan-Meier curves by high and low SUVmax value. A local recurrence-free survival, B regional recurrence-free survival for patients stratified by the cutoff value of SUVmax value; C overall survival

NTR and prognosis in high-risk group

We further analyzed the high-risk group patients. ROC curve analysis revealed that the optimal cutoff value of the NTR for predicting DMFS was 0.23 in the LH group, 2.35 in the HL group, and 1.29 in the HH group. The high-risk group patients were again reorganized into a low NTR group (NTR ≤ 0.23, ≤ 2.35, and ≤ 1.29 in the LH, HL, and HH groups, respectively) and a high NTR group (NTR > 0.23, > 2.35, and > 1.29 in the LH, HL, and HH groups, respectively).

The 3-year LRFS, RRFS, DMFS, PFS, and OS rates in the low NTR vs. high NTR group were 94.1% vs. 86.0% (P = 0.082), 95.9% vs. 92.5% (P = 0.160), 92.3% vs. 80.6% (P = 0.009), 84.0% vs. 67.7% (P = 0.011), and 95.9 vs. 89.2% (P = 0.002), respectively. The OS, PFS, and DMFS rates were significantly worse in the high NTR group than in the low NTR group (all P < 0.005; Fig. 4).

Fig. 4
figure 4

Kaplan-Meier by high NTR and low NTR group in high-risk patients. A local recurrence-free survival, B regional recurrence-free survival, C distant metastasis-free survival, D progression-free survival, and E overall survival

Multivariate analysis with other known prognostic factors, including tumor stage (T stage), nodal stage (N stage), and age, revealed that the NTR was an independent prognostic factor for DMFS (HR = 2.037, 95% CI: 1.039–3.992, P = 0.038), PFS (HR = 1.636, 95% CI: 1.021–2.621, P = 0.041), and OS (HR = 2.543, 95% CI: 1.214–5.325, P = 0.013). The results of the multivariate analysis are summarized in Table 2.

Table 2 Multivariate analysis of LRFS, RRFS, DMFS, PFS and OS for NTR

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated in a credit line to the data.


This article is autogenerated using RSS feeds and has not been created or edited by OA JF.

Click here for Source link (