Study selection
Figure 1 shows the flowchart for study selection. A total of 668 potential records were initially identified by searching the electronic databases (Figure 1). After the initial exclusion of the duplicates, reviews, conference abstracts, and letters (n = 421), and further exclusion of 214 and 6 records after assessing the titles and abstracts, respectively, 27 studies remained for full-text reading. No additional articles were identified through manual search from the reference list of the included studies. The kappa coefficients for the study selection by assessing the titles and abstracts were 0.81 and 0.83, respectively. Eighteen studies were subsequently excluded due to the specific reasons (Table 1) [9–18, 32–39], leaving 9 studies for the further systematic review and analysis [19–27].
Table 1
Numbers and reasons for the excluded articles after full-text review
Reasons to exclude the articles after the full-text review | Numbers of studies | Numbers in the references list |
Patients treated with other than IMRT | 3 | [9–11] |
HRs of CNN for DMFS was not performed or unable to calculate | 7 | [12–18] |
Survival analysis was not performed for CNN | 2 | [32, 33] |
CNN + necrotic primary tumour as one variable was assessed | 1 | [34] |
Studies conducted by the same institution with overlapping patient recruitment periods and analysis performed for the same CNN pattern | 5 | [35–39] |
IMRT = intensity-modulated radiotherapy, DMFS =distant metastases free survival, HR =hazard ratio, CNN = cervical nodal necrosis |
Characteristics of the eligible studies
Characteristics of the included studies are shown in Table 2. The included studies were conducted by 5 institutions and the CNN were all evaluated on MRI. Three conducted by the same institution were included because these studies performed analysis to evaluate prognostic values of different CNN patterns [25–27]. Analysis was performed to evaluate prognostic values of 6 CNN patterns, which included (1) CNN presence/absence in any nodal group (n =6) [19–24], (2) CNN presence/absence in retropharyngeal nodes (RPNs) (n =1) [25], (3) CNN grades (n =1) [26], (4) CNN laterality (n =1) [27], (5) total CNN volume [23], and (6) maximum percentage of nodal necrotic volume of one single node (necrosis%) [23]. According to the NOS criteria, the quality of the eligible studies ranged from 7 to 9 with a median score of 8 (Table 3).
Table 2
Characteristics of the eligible articles
First author | Year of Publication | City | Patient recruitment period | Total patients | Patient treated with chemotherapy | Patient with nodes | Patient with CNN | Imaging modality | Numbers of undifferentiated + non-keratinising types | AJCC /UICC Edition | Median follow-up time (range) (months) | Patient group for analysis |
CNN presence/absence in any nodal group |
Li [19] | 2013 | Guangzhou | 2003-2007 | 749 | 535 | 565 | 142 | MRI | 744 | 7th | 60.7 (3-104) | N+ group |
Zhang [20] | 2017 | Guangzhou | 2009-2012 | 1302 | 1193 | 1302 | 448 | MRI | 1294 | 7th | 47.8 (1.3-75.3) | N+ group |
Zhou [21] | 2018 | Shanghai | 2010-2011 | 354 | 300 | 320 | 143 | MRI | 353 | 7th | 63 (Not mentioned) | All |
Feng [22] | 2019 | Hangzhou | 2007-2012 | 616 | 601 | 616 | 235 | MRI | 612 | 8th | 62.6 (3.4 -119) | N+ group |
Ai [23] | 2019 | Hong Kong | 2005-2012 | 546 | 382 | 404 | 153 | MRI | 544 | 8th | 82.3 (3.2-150) | N+ group |
Xu [24] | 2021 | Xi’an | 2006-2018 | 792 | 744 | 687 | 501 | MRI | 789 | 7th | 46.2 (1.3-130) | All |
CNN presence/absence in retropharyngeal nodes |
Tang [25] | 2014 | Guangzhou | 2003-2007 | 749 | 535 | 565 | 64/484 (RPN+) | MRI | 744 | 7th | 81 (3-127) | N+ group |
CNN grades |
Zhang [26] | 2017 | Guangzhou | 2009-2012 | 1423 | 1310 | 1423 | Grade 1: 213 Grade 2: 257 | MRI | 1415 | 7th | 48.6 (1.3-76) | N+ group |
CNN laterality |
Xie [27] | 2020 | Guangzhou | 2010-2013 | 733 | 634 | 559 | No/unilateral CNN: 692 Bilateral CNN: 41 | MRI | 728 | 8th | 62 (1.4-83.2) | All |
CNN =cervical nodal necrosis, AJCC/UICC =American Joint of Cancer Committee/ Union for International Cancer Control, N+group =patients with metastatic nodes, RPN =retropharyngeal node, MRI =magnetic resonance imaging |
Table 3
The Newcastle-Ottawa Scale(NOS) quality assessment of the eligible studies in the meta-analysis
First author | Year of publication | Selection | Comparability | Outcome | Total score |
Representativeness of the exposed cohort (0 - 1) | Selection of the non-exposed cohort (0 -1) | Ascertainment of exposure (0 -1) | Demonstration that outcome of interest was not present at start of study (0 -1) | Comparability of cohorts on the basis of the design or analysis (0 -2) | Assessment of outcome (0 -1) | Was follow-up long enough for outcomes to occur (0 -1) | Adequacy of follow up of cohorts (0 -1) |
CNN presence/absence in any nodal group |
Li [19] | 2013 | 1 | 1 | 0 | 1 | 2 | 1 | 1 | 1 | 8 |
Zhang [20] | 2017 | 1 | 1 | 1 | 1 | 2 | 0 | 1 | 1 | 8 |
Zhou [21] | 2018 | 1 | 1 | 1 | 1 | 2 | 0 | 1 | 1 | 8 |
Feng [22] | 2019 | 1 | 1 | 1 | 1 | 2 | 0 | 1 | 1 | 8 |
Ai [23] | 2019 | 1 | 1 | 1 | 1 | 2 | 1 | 1 | 1 | 9 |
Xu [24] | 2021 | 1 | 1 | 0 | 1 | 2 | 0 | 1 | 1 | 7 |
CNN presence/absence in retropharyngeal nodes |
Tang [25] | 2020 | 1 | 1 | 0 | 1 | 2 | 1 | 1 | 1 | 8 |
CNN grades |
Zhang [26] | 2021 | 1 | 1 | 1 | 1 | 2 | 1 | 1 | 1 | 9 |
CNN laterality |
Xie [27] | 2021 | 1 | 1 | 1 | 1 | 2 | 1 | 1 | 1 | 9 |
Three studies were excluded from the meta-analysis due to the great heterogeneity of the analysed CNN patterns or the limited studies for evaluating each of the CNN patterns. Meta-analysis therefore was only eligible to perform in 6 studies which evaluated the prognostic value of CNN presence/absence in any nodal group [19–24]. The eligible studies for meta-analysis were published over 9 years (2013[19]– 2021 [24]) (Table 2). Patient recruitment ranged from 2003 to 2018 with the follow-up periods ranged from 1.3 to 150 months (Table 2). The total patient number extracted from these studies was 4359 (range 354 –1302) with ages ranging from 12 to 90 years, of which 3216 (73.8%) were male and 1143 (26.2%) were female. Metastatic cervical nodes were observed in 3894 patients, of which CNN was observed in 1622/3894 (41.7%) patients (Table 2). Over 99% of patients had undifferentiated carcinoma or non-keratinising carcinoma (Table 2).
Prognostic value of the CNN presence/absence
Of the 6 studies that are eligible for meta-analysis, 4 studies performed the survival analysis in patients with metastatic nodes (N+ group), and 2 in all patients (Table 2). The unadjusted and adjusted HRs of CNN were reported for DMFS in 5 and 5 studies respectively, for DFS in 3 and 4 studies, respectively, and for OS in 3 and 4 studies, respectively (Figure 2-4).
The pooled unadjusted HRs showed that the presence of CNN predicted poor DMFS (HR = 1.89, 95%CI = 1.72 – 2.08, I2 = 44.8%, p = 0.123) (Figure 2), DFS (HR = 1.57, 95%CI = 1.08 – 2.26, I2 = 71.1%, p = 0.063) (Figure 3), and OS (HR = 1.87, 95%CI = 1.69 – 2.06, I2 = 9.1%, p = 0.333) (Figure 4). The pooled adjusted HRs also showed the consistent results for DMFS (HR = 1.34, 95%CI = 1.17 – 1.54, I2 = 0.0%, p = 0.426) (Figure 2), DFS (HR = 1.30, 95%CI = 1.08 – 1.56, I2 = 12.7%, p = 0.318) (Figure 3), and OS (HR = 1.61, 95%CI = 1.27 – 2.04, I2 = 0.4%, p = 0.390) (Figure 4). The sensitivity tests of the unadjusted and adjusted HRs for the survival endpoints in the meta-analysis are shown in Figure 5. Subgroup meta-analysis was performed for the adjusted HRs to further evaluate the prognostic value of CNN presence/absence in patients with metastatic nodes (N+ group). Results showed that the presence of CNN predicted poor DMFS (HR = 1.56, 95%CI = 1.25 – 1.95, I2 = 0.0%, p = 0.837) (Figure 2), DFS (HR = 1.43, 95%CI = 1.05 – 1.95, I2 = 40.1%, p = 0.196) (Figure 3), and OS (HR = 1.55, 95%CI = 1.21 – 2.00, I2 = 12.3%, p = 0.320) (Figure 4).
The Begg’s and Egger’s tests showed that no potential publication bias were observed in the meta-analysis (Begg’s, p = 0.211 to >0.999; Egger’s test, p =0.132 to 0.905) except for that of the unadjusted HRs for DMFS (Egger’s test, p = 0.014).
Prognostic values of other CNN patterns
Tang et al[25] evaluated the prognostic value of the presence of CNN in the retropharyngeal nodes (RPNs) in patients with metastatic nodes (N+ group) showing the presence of CNN independently predicted poor DMFS (HR = 1.75, 95%CI =1.10 – 2.79), and DFS (HR = 1.80, 95%CI =1.21 – 2.65). Zhang et al[26] classified patients with nodes into 3 grades of necrosis (grade 0: no necrotic area; grade 1: any node with necrotic area of ≤33%; and grade 2: any node with necrotic area of > 33%) and added the necrosis grades as the continuous variable to the survival analysis. Results showed that patients with higher necrosis grades independently predicted poorer DMFS (HR = 1.36, 95%CI =1.14 – 1.63), DFS (HR = 1.38, 95%CI =1.21 – 1.59), and OS (HR = 1.36, 95%CI =1.13 – 2.45). Xie et al[27] reported patients with bilateral CNN independently predicted poorer DMFS (HR = 2.10, 95%CI =1.10 – 2.40) compared to those with unilateral CNN or without CNN. The study conducted by Ai et al.[23] also quantitatively evaluated the total CNN volume and necrosis% showing necrosis%, but not the total CNN volume, was a factor for predicting DMFS (HR = 3.03, 95%CI =1.242–7.397) and OS (HR = 3.09, 95%CI =1.482–6.431); while necrosis% was not an independent factor to predict outcome when other confounding factors were added to the multivariate analysis.