The role of concurrent chemoradiotherapy in the treatment of locoregionally advanced nasopharyngeal carcinoma among endemic population: a meta-analysis of the phase iii randomized trials

Background The main objective of this meta-analysis was to determine the clinical benefit of concurrent chemoradiotherapy (CCRT) compared with radiation alone (RT) in the treatment of nasopharyngeal carcinoma (NPC) patients in endemic geographic areas. Methods Using a prospective meta-analysis protocol, two independent investigators reviewed the publications and extracted the data. Published randomized controlled trials (RCTs) in which patients with NPC in endemic areas were randomly assigned to receive CCRT or RT alone were included. Results Seven trials (totally 1608 patients) were eligible. Risk ratios (RRs) of 0.63 (95% CI, 0.50 to 0.80), 0.76 (95% CI, 0.61 to 0.93) and 0.74 (95% CI, 0.62 to 0.89) were observed for 2, 3 and 5 years OS respectively in favor of the CCRT group. The RRs were larger than that detected in the previously reported meta-analyses (including both endemic and non-endemic), indicating that the relative benefit of survival was smaller than what considered before. Conclusions This is the first meta-analysis of CCRT vs. RT alone in NPC treatment which included studies only done in endemic area. The results confirmed that CCRT was more beneficial compared with RT alone. However, the relative benefit of CCRT in endemic population might be less than that from previous meta-analyses.


Background
Nasopharyngeal carcinoma (NPC) is a common malignant disease of the head and neck with a high prevalence in Southern China and Southeast Asia. It is different from other head and neck cancers because of unique epidemiology, natural behavior and therapeutic considerations.
NPC is both a radiosensitive and chemosensitive tumor. Since the publication of the results of a multicentre randomized trial conducted in North America (Intergroup study 0099) [1], concurrent chemoradiotherapy (CCRT) has been accepted as standard in the treatment of patients with stage III and IV NPC gradually. However, the major concern remains in extrapolating the findings of the intergroup study to patient groups in the Asian context, where NPC is endemic. Several metaanalyses and a pooled data analysis [2][3][4][5] had shown an improvement of survival in NPC patients who received chemotherapy and radiotherapy (CR+RT) versus those received radiotherapy alone (RT). Unfortunately, it still remains unclear regarding the benefit of CCRT especially for endemic population in the previously published meta-analyses. This fact is all these meta-analyses included heterogeneous histological mix of patients, limited number of studies published, or complexity of study design (CCRT with or without adjuvant or neoadjuvant chemotherapy versus RT alone). In contrast, a number of clinical studies [6][7][8][9][10][11][12][13][14][15][16][17][18] mainly focus on the additional value of CCRT from endemic areas has been published in recent years.
To gain a better understanding of the potential benefit of CCRT in endemic population, we undertook a metaanalysis that pooled data from all published Phase III randomized controlled trials (RCTs) done in endemic areas focusing on the impact of CCRT comparing with RT alone on patients with locally advanced NPC. To our knowledge, this is the first meta-analysis that included only those randomized trials done in endemic areas to date. The pooled analysis of largest cohort (1608 patients) should provide a clearer understanding of the impact of CCRT on the natural history of this disease.

Methods
A prospective meta-analysis protocol including study aim, study selection criteria, literature search strategy, quality control of literature and statistical procedures was developed. The primary aim of present analysis was designed to evaluate how the CCRT influenced survival at 2, 3 and 5 years after treatment compared with RT alone in endemic area patients with locally advanced NPC. More specifically, the analysis was designed to examine the difference in patterns of failure (locoregional recurrence, distant metastasis) in CCRT and RT alone treatment group. In addition to the main metaanalysis, we also compared the survival difference between the CCRT with/without adjuvant chemotherapy (AC) and RT alone.

Study Criteria
The selection criteria for eligible studies in this metaanalysis included published randomized controlled trials done in endemic area recruiting NPC patients of Asian origin. Patients were randomly assigned to receive radiotherapy alone or concurrent chemotherapy combined with radiotherapy. Patients receiving concurrent chemotherapy plus some form of adjuvant chemotherapy in addition to radiotherapy were also included in this analysis. The 1997 UICC TNM staging system was used for the staging of the primary tumor. CT or MRI was used as the main evaluation method and adequate doses of radiotherapy was given in both arms equivalent to at least 64Gy, with conventional fractionation to the primary lesion. Overall survival (OS) was the primary outcome measure for measuring the effect of treatment.

Literature Search Strategy
The meta-analysis aimed to include all the relevant published trials done in endemic areas. To conduct a search of the published literatures, multiple search tools by two independent investigators were used: 1.Computerized bibliographic databases: Electronic databases (MED-LINE, CANCERLIT, and EMBASE) were searched with the medical headlines such as Nasopharyngeal carcinoma, Concurrent chemoradiotherapy, radiotherapy, and randomized controlled trials, to identify potentially eligible trials. 2. Journal manual search and reference lists: The computer search was supplemented with manual search of reference lists of all available review articles, primary studies, abstracts from meetings, and bibliographies of books. 3. Conference proceedings of ASCO, ESMO/ECCO, ASTRO and ESTRO.
11 clinical trials including 2252 patients were initially identified. Among them, 4 trials were secondarily considered ineligible for different reasons ( Figure 1) [1,[6][7][8][9][10][11][12][13][14][15][16][17][18]. In total, 4 trials were excluded. The trial reported by Al-Sarraf et al [1] was excluded because this randomized trial was conducted in North America. The trials by Chua DT et al [6] and Mizowaki et al [7] were excluded in this meta-analysis for these trials not being randomized trials. The trial by Lin et al [8,9] was excluded for lack of samples size calculation and introduction of randomization method. It was also excluded by another earlier meta-analysis by Bertrand Baujat [5] for not meeting the eligibility criterion of unpredictable treatment assignment.

Application of literature quality
The trials were reviewed using a list of predefined pertinent issues that concerned the characteristics of patients and treatments. To assess the methodological quality of RCTs, we have examined the statistical design, the randomization process, the quality control process, the dropout rate and if potential bias was taken into account. All analyses followed the intention-to-treat principle. This meta-analysis was performed according to preferred reporting items for systematic reviews and meta-analyses -The PRISMA statement [19].

Method of data retrieves
Two independent investigators reviewed the publications and extracted the data. The following information was extracted from each article: 1. Basic information from papers such as, year of publication, journal name, and author name etc. 2. Characteristics of patients such as: age, sex, pathological types, stage, and study duration. 3. Information of study designation such as: sample size per group, study design, randomization scheme, inclusion criteria, and type of end point used. 4. Information of treatment such as: treatment modality, dose of RT, response rate of treatment, numbers of death, locoregional recurrence, distant metastasis, withdrawals, and so on. Available information was extracted and recorded to a data collection form and entered into electronic database.

Statistical Analysis
The primary end point of this meta-analysis was OS, defined as the time from random assignment to death. Secondary end points were the incidence of local and/or regional recurrence (LRR) and distant metastasis (DM). Extraction of summary statistics from the Kaplan-Meier curve was performed according to standard methods for survival endpoints proposed by Parmar et al [20]. Standard techniques for meta-analysis were used [21].
Results were expressed as relative risk (RR) with 95% confidence intervals (CIs). The RR of survival at 2, 3, and 5 years and RR of LRR or DM at 3 years were compared between the CCRT and RT alone groups. RR less than 1 indicated improved survival for the combinedmodality treatment compared with radiotherapy alone. When the span of the 95% CI given did not include 1, the result was statistically significant. Before estimation of a RR, a statistical test for homogeneity was performed. A Dersimonian and Laird random effects model [22] was used in cases in which statistically significant heterogeneity between studies likely existed. If no significant heterogeneity was found, a fixed-effects model was used to calculate pooled RR and 95% CIs. All analyses were conducted using Review Manager Version 5.0.24 (Revman; the Cochrane Collaboration; Oxford, England).

Study identification and eligibility
After the selection procedure ( Figure 1), 7 trials were considered eligible [10][11][12][13][14][15][16][17][18]. The majority of patients were included after 1990. The characteristics of these studies are listed in Table 1. Kwong's trial is a factorially designed study to test the efficacy of CCRT and adjuvant chemotherapy (AC) independently [10]. Patients were divided into four treatment groups: Group A, RT alone; Group B, CCRT; Group C, RT and AC; Group D, CCRT and AC. As in our analysis, we only took those randomized studies, which included comparison between CCRT vs. RT or CCRT +AC vs. RT. We did not include the patients in group C for the analysis. Subsequently for OS and overall locoregional and distance failure, group A was compared with group B and D. For subgroup analysis with pure CCRT vs. RT, groups A and B were compared and with adjuvant groups A and D were compared.

Overall survival
Data regarding the OS of all the 7 trials [10][11][12][13][14][15][16][17][18] were available. The trial by Kwong et al [10] was not included in 2 years and 5 years OS calculations due to insufficient data. The trial by Chen et al [16] was not included in 3 years and 5 years OS calculations for too early to get 3 and 5-year data. Our own trial (Zhang et al [14]) was not included in 5 years OS calculations also for too early to get 5 years data. When 2-year overall survival rates were calculated, it showed significant benefit in favor of the CCRT treatment with RR of 0.63 (95% CI, 0.50 to 0.80). 3-year OS also showed significant benefit in favor of the CCRT treatment with RR of 0.76 (95% CI, 0.61 to 0.93). 5 years OS was significantly better in favor of the CCRT treatment groups with an RR of 0.74 (95% CI, 0.62 to 0.89) ( Figure.2).
For the second part of the analysis, we found that CCRT without AC was better than RT alone for 3 years OS. CCRT vs. RT showed RR of 0.66 (95% CI, 0.48 to 0.92) for 3 years OS, likewise CCRT+AC vs. RT showed RR of 0.83 (95% CI, 0.63 to 1.09) ( Figure 3).

Locoregional Recurrence
Data regarding the absolute number of locoregional recurrence rate (LRR) for 3 years were available in the 6 studies [10][11][12][13][14][15]. A significant overall benefit in favor of the addition of chemotherapy was found with RR of 0.67 (95% CI, 0.49 to 0.91). The failure rate of locoregional control of CCRT group seems better compared to the RT alone group, the difference was significant ( Figure. 4).

Distant Metastasis
Data regarding the absolute number of distant metastasis rate (DMR) in 3 years were provided in 6 trial reports [10][11][12][13][14][15]. A significant overall benefit in favor of the addition of chemotherapy was found with RR of 0.71 (95% CI, 0.58 to 0.88) (Figure. The RR for DMR of CCRT vs. RT alone [10,11,14] and DMR of CCRT +AC vs. RT alone [10,12,13,15] were also calculated. The CCRT group had the RR of 0.71 (95% CI, 0.51 to 0.99) compared with RT alone and the group with CCRT plus AC showed the RR of 0.71 (95% CI, 0.54 to 0.92) compared with RT alone (Figure. 5). This meta-analysis was designed to directly address the additional effect of chemotherapy concurrently combined with radiotherapy (CCRT) in endemic NPC population. These results suggested that the superior survival observed with CCRT compared with RT alone may be related significantly with improvement in the risk of distant metastasis.
Our analysis differed a little from the results of previous meta-analyses. The RR of 3 years OS (RR = 0.76)

Study or Subgroup
Chan (11,25) Kwong (10) Lee 9901 (13,18) Lee 9902 (15,17) Wee (12) Zhang (14) Total (95% CI)   was larger than that detected in the other meta-analyses (RR = 0.16-0.60) [2,3,5], indicating that the benefit was smaller than anticipated ( Table 2). The possible explanation for these differences is high proportion of patients with WHO type I histology in previous meta-analyses. In the current meta-analysis, almost all the cases were histologically proven NPC. More than 99.69% of these cases belonged to the WHO Grade II and Grade III subtypes. Only about 0.31% of the NPCs belonged to the WHO Grade I subtype. Whereas in other similar meta-analyses done in the past [2,3,5], significantly more percentage of patients with WHO Grade I type of tumor were included. As Grade I type of NPC is similar to squamous cell carcinoma of head and neck, which is more resistant to RT than Grade II and III type of NPC, patients with Grade I NPC subtypes may benefit mostly from CCRT [3,10]. But in endemic areas, most of the patients are Grade II and III type tumors, which are more sensitive to RT. In the current meta-analysis, we found that the contribution of survival benefit of CCRT

Total events
Heterogeneity: Chi² = 4.53, df = 2 (P = 0.10); I² = 56% Test for overall effect: Z = 2.50 (P = 0.01)    mainly came from improvement of distant failure. This result also implied that NPCs in endemic areas are sensitive to RT. On the other hand, the RT methods are more aggressive in endemic areas than in other areas [1,10]. Therefore the margin of benefit potentially gained with additional chemotherapy may be reduced [12]. The finding that the RT alone group showed better 5-year survival rate in our analysis than in that reported by the intergroup study [1] (62% vs 21%) may also support this hypothesis. The current meta-analysis result indicated that CCRT was still the most effective treatment modality for the improvement of overall survival, but the exact magnitude of treatment effect of CCRT in endemic areas might be less than that previously reported by other meta-analyses. Further studies should be carried on exploring less toxic chemotherapy regimen for CCRT.

Events
In this analysis, we also tried to find if there is any additional benefit for the patients receiving CCRT plus some kind of AC. Sub-group analyses showed that both

Total events
Heterogeneity: Chi² = 4.44, df = 2 (P = 0.11); I² = 55% Test for overall effect: Z = 1.30 (P = 0.19)   locoregional recurrences and distant failure were improved in CCRT+AC arms compared with RT alone. Pooling the results of these studies, the RR of death with CCRT+AC was not significant (figure 3, P = 0.18). This lack of survival difference might be due to increased mortality related to the toxicities of chemotherapy, and possibly successful salvage after relapse. Increase of non-cancer deaths due to treatment-related, incidental, and unknown causes might have also narrowed the actual magnitude of survival gain [13,15,17,18]. The role of adjuvant chemotherapy remains to be addressed by additional studies.
There was an indication from a previous trial by Lin et al [9] that the benefit of AC was associated with so called high-risk patients who met at least one of the following criteria: (1) nodal size >6 cm, (2) supraclavicular node metastases, (3) 1992 AJCC stage T4N2, (4) multiple neck node metastases with 1 node >4 cm. As our analysis did not include enough individual data, we could not analyze the effect of CCRT and AC vs. RT

Study or Subgroup
Chan (11,25) Kwong (10) Wee (12) Zhang (14) Lee 9901 (13,18) Lee 9902 (15,17) Total (95% CI)   Figure 5 Three years Distant metastasis rates with RT alone versus CCRT with and/or without AC. Forest plot of the risk ratio (RR) for 3 years distant metastasis with RT alone versus CCRT with and/or without AC. The estimate of the RR of each individual trial corresponds to the middle of squares and the horizontal line gives 95% CI. The closed diamond shows the overall RR with its 95%CI. RR less than 1 indicated improved survival for the CCRT compared with RT alone. Kwong_AB: In Kwong's trial, for subgroup analysis with pure CCRT vs. RT, groups A and B were compared. Kwong_AD: In Kwong's trial, for subgroup analysis with adjuvant groups, A and D were compared.
alone for the high-risk patients. We recommend that the future trials should be focused on the high risk patients.
One of the shortcomings of our meta-analysis is that all information came from published data instead of individual patient data, which might result in two sources of bias -publication bias and selection bias. We used funnel plot to estimate the publication bias. If the funnel plot is not symmetrical or not integrated then it suggests that the result may be biased. So we tested it by using the linear regression model proposed by Egger [23,24]. In our studies the funnel plot was symmetrical, suggesting that publication bias was not significant. To avoid the selection bias, two independent investigators reviewed the publications and extracted the data. The heterogeneity between the individual studies was also evaluated.

Conclusions
In conclusion, our meta-analysis based on published trials in endemic areas showed that the CCRT was the most effective treatment modality for the improvement of overall survival in locally advanced NPC. However, the relative benefit of CCRT in endemic population might be very different from previously published metaanalyses. In the future, treatment of NPC should be individualized, according to recognized prognostic factors, while recognizing the results of randomized trials of induction and concurrent CCRT.