Chemotherapy as an adjunct to radiotherapy in locally advanced nasopharyngeal carcinoma

Abstract

Background

A previous meta‐analysis investigated the role of chemotherapy in head and neck locally advanced carcinoma. This work had not been performed on nasopharyngeal carcinoma.

Objectives

The aim of the project was to study the effect of adding chemotherapy to radiotherapy on overall survival (OS) and event‐free survival (EFS) in patients with nasopharyngeal carcinoma.

Search methods

We searched MEDLINE (1966 to October 2003), EMBASE (1980 to October 2003) and the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library, Issue 3, 2003) and trial registers. Handsearches of meeting abstracts, references in review articles and of the Chinese medical literature were carried out. Experts and pharmaceutical companies were asked to identify trials.

Selection criteria

Randomised trials comparing chemotherapy plus radiotherapy to radiotherapy alone in locally advanced nasopharyngeal carcinoma were included.

Data collection and analysis

The meta‐analysis was based on updated individual patient data. The log rank test, stratified by trial, was used for comparisons and the hazard ratios (HR) of death and failure (loco‐regional/distant failure or death) were calculated.

Main results

Eight trials with 1753 patients were included. One trial with a 2 x 2 design was counted twice in the analysis. The analysis was performed including 11 comparisons based on 1975 patients. The median follow up was six years. The pooled hazard ratio of death was 0.82 (95% confidence interval (CI) 0.71 to 0.95; P = 0.006) corresponding to an absolute survival benefit of 6% at five years from chemotherapy (from 56% to 62%). The pooled hazard ratio of tumour failure or death was 0.76 (95% CI 0.67 to 0.86; P < 0.00001) corresponding to an absolute event‐free survival benefit of 10% at five years from chemotherapy (from 42% to 52%). A significant interaction was observed between chemotherapy timings and overall survival (P = 0.005), explaining the heterogeneity observed in the treatment effect (P = 0.03) with the highest benefit from concomitant chemotherapy.

Authors' conclusions

Chemotherapy led to a small but significant benefit for overall survival and event‐free survival. This benefit was essentially observed when chemotherapy was administered concomitantly with radiotherapy.

Plain language summary

Chemotherapy as an adjunct to radiotherapy in locally advanced nasopharyngeal carcinoma

Eight trials (1753 patients) met the criteria for inclusion in this review. The addition of chemotherapy to standard radiotherapy provides a small but significant benefit in patients with nasopharyngeal cancer, especially when chemotherapy is administered at the same time as radiotherapy. The role of chemotherapy given before or after the radiotherapy is more questionable.

Background

Nasopharyngeal carcinoma (NPC) is pathologically, epidemiologically and clinically distinct from other head and neck cancers (Ali 2000; Vokes 1997). Nasopharyngeal carcinoma is rare in the USA and Western Europe and in these areas, the frequency of squamous cell carcinoma (World Health Organization (WHO) type 1) is about 25%, which is markedly higher than in endemic areas. Endemic areas include Southern China, Southeast Asia, Middle East, North Africa, Alaska and Greenland. In these areas there is a strong correlation between Epstein‐Barr virus latency and nasopharyngeal carcinoma. Most patients have a poorly or undifferentiated carcinoma (WHO type 2 or 3)and present with locally advanced stage disease. Nodal involvement and bilateral nodal disease are more frequently observed with nasopharyngeal carcinoma than with other head and neck cancers.

Nasopharyngeal carcinoma is commonly treated with radiation therapy and chemotherapy (Ali 2000). Radiotherapy at a dose of 65 to 75 Gray (Gy) over six to seven weeks is the standard treatment. Overall survival (OS) at five years ranges from 32% to 52% in large series of patients with locally advanced disease treated with radiotherapy alone (Ali 2000).

Chemotherapy has been proposed for locally advanced nasopharyngeal carcinoma to improve survival (Ali 2000; Vokes 1997). In spite of 11 randomised trials comparing radiotherapy to radiotherapy plus chemotherapy in the English language literature, the magnitude of the effect of chemotherapy on survival is not well‐established. Overall survival was a main endpoint in all of these trials excepted one, but only four trials showed a beneficial effect on event‐free survival, and two of them showed a beneficial effect on overall survival. Underpowered trials could account for the inconstancy of the benefit for survival, which was the case in the previously published Meta‐Analysis of Chemotherapy in Head and Neck Cancer (MACH‐NC) (Pignon 2000).

Objectives

The aim of the Meta‐Analysis of Chemotherapy in Nasopharynx Carcinoma (MAC‐NPC) was to assess the impact of adding chemotherapy to radiotherapy on overall survival.

Methods

Criteria for considering studies for this review

Types of studies

Trials were eligible if accrual had been completed before 31st December 2001. Each trial had to be randomised in a manner precluding prior knowledge of treatment assignment.

Types of participants

Previously untreated patients with non‐metastatic nasopharyngeal carcinoma (WHO type 1, 2 or 3).

Types of interventions

Trials were eligible if radiotherapy plus chemotherapy had been compared to radiotherapy alone as a potentially curative loco‐regional treatment.

Types of outcome measures

The main endpoint was overall survival which was evaluated from the time of randomisation until death, whatever the cause. The survival of living patients was taken into account in the analyses until the date of their last follow up. The secondary endpoint was event‐free survival (EFS), i.e. the time from randomisation until the first event including loco‐regional/distant failure or death.

Search methods for identification of studies

Published and unpublished trials performed between 1975 and 2001 were included. Computerised searches of MEDLINE and EMBASE were supplemented with handsearches of meeting abstracts and references in review articles. Trial registers managed by the National Cancer Institute (PDQ®, ClinProt) were consulted. Experts, pharmaceutical companies and all trialists who took part in the meta‐analysis were asked to identify trials. This first search ended in May 2001. A second‐line search was done by Dr Thephamongkhol in October 2003: The literature was searched using MEDLINE (1966 through October 2003), EMBASE (1980 through October 2003), the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library Issue 3, 2003), the Physician Data Query (PDQ®) database, the Canadian Medical Association (CMA) Infobase, the National Guideline Clearinghouse, and abstracts published in the proceedings of the meetings of the American Society of Clinical Oncology (1997 to 2003), the American Society for Therapeutic Radiology and Oncology (1992 to 2002), the Asian Clinical Oncology Society (2001), the International Congress of Radiation Oncology (1997 and 2001), the European Society of Therapeutic Radiology and Oncology (1992, 1994, 1996, 1998, 2000 and 2002), and the European Society for Medical Oncology (2000 and 2002). Article bibliographies and personal files were also searched to October 2003 for evidence relevant to this practice guideline report. The literature search combined nasopharyngeal disease specific terms (nasopharyngeal neoplasms/ or nasopharyn:.mp. or nasopharyngeal.tw.) with treatment specific terms (drug therapy/ or chemotherapy/ or chemotherapy.tw. or radiochemotherapy.mp. or chemoradiotherapy.mp.) and search specific terms for the following study designs: practice guidelines, systematic reviews, meta‐analyses, reviews, randomised controlled trials and clinical trials (Thephamongkhol 2004). A second‐line search of Chinese and occidental literature was carried out by Dr Thephamongkhol in October 2003.

Data collection and analysis

The meta‐analysis was based on individual patient data (Pignon 2001) and used a methodology similar to that used in the MACH‐NC study (Pignon 2000) and the Prophylactic Cranial Irradiation Overview (Auperin 1999).

All data were checked for internal consistency and compared with the trial protocol and published reports. Ranges were checked and extremes were verified with the trialists. Each trial was re‐analysed and the analyses were sent to the trialists for review.

The data collected for each patient included: age, gender, WHO performance status (or equivalent), histology (WHO), TNM (Tumour, Node, Metastasis) staging, treatment allocated, date of randomisation, cause of death, date of loco‐regional failure, date of distant failure, date and type of second primary, exclusion (yes/no) from the trial analysis and the reason for exclusion, at least one cycle of chemotherapy received (yes/no). As different TNM staging classifications were used in the publications, authors who used the Ho classification in their publication were requested to provide, if possible, American Joint Committee on Cancer (AJCC)/International Union Against Cancer (UICC) staging. The last follow‐up and survival status were updated as far as possible, as compared with the published analyses. Early deaths (i.e. within three months) were also studied.

All analyses were on an intention‐to‐treat basis that is, all randomised patients were included in the analyses according to the allocated treatment, irrespective of whether they received the treatment or whether they were excluded from the investigator's original analysis. Median follow up was quantified using the potential follow‐up method (Schemper 1996). Survival analyses were stratified by trial, and the log rank Observed minus Expected number of deaths (O‐E) and its variance were used to calculate individual and overall pooled hazard ratios with a fixed‐effect model (EBCTCG 1990). The absolute differences in the two and five year survival rates were calculated using the pooled hazard ratio and the assumption of proportional hazards was used to calculate survival rates at two and five years using the control radiotherapy group (Stewart 1993). Non‐stratified Kaplan‐Meier survival curves are presented for the control and the treatment groups. Chi‐square heterogeneity tests were used to test for statistical heterogeneity among trials (Higgins 2002). The percentage of variability in the estimates of the treatment effect due to heterogeneity between studies rather than sampling errors was estimated by I² statistics (Higgins 2002). A limited number of comparisons, planned in the meta‐analysis protocol, were done between subsets of trials and subgroups of patients, and a test for interaction or a test for trend was performed to look for any significant variation in the treatment effect among these subgroups/subsets (NSCLCCG 1995). All P‐values are two‐sided.

The meta‐analysis protocol (MAC‐NPC 2004) specified that covariates to be considered would be: Age in three categories as in the MACH‐NC study, but as the patient population was younger in this study, another distribution was used corresponding to the tertiles (40 or less, 41 to 50, 51+), gender, performance status (WHO, 0, 1, 2+), TNM stage and histologic type. Tumour stage was divided into three categories according to the 1997 AJCC/UICC classification (T1, T2, T3 to 4) because five trials used AJCC/UICC classifications before 1997, one trial used Ho's classification, and the other two used the 1997 AJCC/UICC classification, thus rendering the distinction between T3 and T4 patients impossible. Node stages had to be divided into three categories (N0, N1 to 2, N3) for the same reason. We could not use the data from the trial that used Ho's classification (Chan 1995) for this covariate because nodes are classified topographically in Ho's staging system. Histologic types were divided into two categories (WHO 1 versus 2 to 3): we had to pool patients with WHO type 2 and 3 carcinoma because data were missing in one trial (only patients with WHO type 2 and 3 carcinoma were included). Trials were grouped according to the timing of chemotherapy: induction (before radiotherapy), concomitant (chemotherapy given concomitantly with radiotherapy) and adjuvant (after radiotherapy). Trials combining induction and adjuvant chemotherapy or concomitant and adjuvant chemotherapy were respectively included in the induction group and in the concomitant group. Trials were also grouped according to the type of chemotherapy: cisplatin + 5‐FU‐based chemotherapy versus other chemotherapy.

Results

Description of studies

Eleven trials including 2722 patients were identified in the English language literature. The data from one trial were lost at the institution (Rossi 1988) (229 patients). We received data from ten trials. Two trials were excluded by the Steering Committee after blind review (740 patients) because they did not fulfil the eligibility criterion of unpredictable treatment assignment (Lin 2003; Ma 2001).

Eighty‐eight comparative trials were found in the Chinese literature (Thephamongkhol 2004) (list available on request). Twelve were selected according to criteria based on a quality score (Thephamongkhol 2004), on size and on the duration of the follow up (1775 patients). We managed to contact nine teams. Data from three trials were lost. Five teams failed to respond despite numerous attempts. Only one database was obtained (Pan 2000) (300 patients) but the trial was also excluded by the Steering Committee after blind review for the same reason as the other two trials.

The database thus included the eight trials (Al‐Sarraf 1998; Chan 1995; Chan 2002; Chi 2002; Chua 1998; Cvitkovic 1996; Hareyama 2002; Kwong 2004a; ), described in the 'Characteristics of included studies' table. All of them have been published as a full paper.

Risk of bias in included studies

All the studies fulfilled our quality criteria. Inclusion curves were compared between both arms to look for randomisation abnomalities. Very few data were missing. Only two trials (299 patients) had a median follow up of less than five years.

Effects of interventions

Population

The eight trials included 1753 patients. Overall, 728 deaths (42%) were observed. The median follow up was six years (range 3 to 9). On the intention‐to‐treat basis, 63 randomised patients who had been excluded from the published studies were included in the present analysis (4%).

One trial had a 2 x 2 design. The 222 patients in this trial were therefore counted twice resulting in a total number of 11 comparisons based on 1975 patients (Kwong 2004a; Kwong 2004b; Kwong 2004c; Kwong 2004d). All trials used conventional radiotherapy, and some patients in one trial (Cvitkovic 1996) received hypofractionated radiotherapy. The doses delivered to the primary tumour site ranged between 65 and 74 Gray (Gy) delivered in 6.5 to 8 weeks. Patients with N0 disease received 50 to 66 Gy. Patients with positive nodes received between 60 and 76 Gy. The 76 Gy dose resulted from an additional boost delivered in case of residual nodes after completion of the treatment in some of the patients in two trials (Chua 1998; Kwong 2004a). Four comparisons (Chan 1995; Chua 1998; Cvitkovic 1996; Hareyama 2002) (830 patients) investigated induction chemotherapy. In one of them (Chan 1995) (77 patients), adjuvant chemotherapy was added to induction chemotherapy. Four comparisons (Al‐Sarraf 1998; Chan 2002; Kwong 2004a; Kwong 2004d) (765 patients) investigated concomitant chemotherapy. In one of them (Al‐Sarraf 1998) (193 patients), adjuvant chemotherapy was added to concomitant chemotherapy. Three comparisons (Chi 2002; Kwong 2004b; Kwong 2004c) (380 patients) investigated adjuvant chemotherapy alone (including a comparison of concomitant chemotherapy versus concomitant chemotherapy + adjuvant chemotherapy). Three comparisons investigated monochemotherapy: UFT (uracil + tegafur) in 222 patients (Kwong 2004a), cisplatin in 350 patients (Chan 2002). The eight other comparisons investigated cisplatin‐based polychemotherapy: cisplatin + 5‐fluorouracil (5‐FU) +/‐ other drugs in 730 patients (Al‐Sarraf 1998; Chan 1995; Chi 2002; Hareyama 2002; Kwong 2004b) and cisplatin + epirubicin +/‐ bleomycin in 673 patients (Chua 1998; Cvitkovic 1996).

Ninety‐three percent of the patients in the radiotherapy + chemotherapy group received at least one cycle of chemotherapy versus 0.1% in the radiotherapy alone group. Patient characteristics are described in Table 1.

1. Patient characteristics by treatment group.

Characteristics	RT + CT	RT
n = 1975	n = 990	n = 985
Gender: male	75%	74%
Age: < 40 or equal	33%	29%
41 to 50	31%	33%
51+	36%	38%
Perform. status (n = 1468)*: 0	52%	50%
1	46%	47%
2	2%	3%
Tumour stage: T1	46%	47%
T2	27%	28%
T3,4	27%	25%
Nodal stage (n = 1 898)**: N0	10%	9%
N1,2	65%	68%
N3	25%	23%
Histology (n = 1636)***: WHO 1	4%	3%
WHO 2	18%	18%
WHO 3	78%	79%
* Data missing from three trials. ** Data missing from one trial, using Ho's classification. *** Data missing from one trial, that did not distinguish between WHO histologic type 2 and 3 but did not include type 1.

Effect of chemotherapy on overall survival

There was a significant 18% reduction in the hazard ratio (HR) of death (P = 0.006) with chemotherapy (HR = 0.82, 95% confidence interval (CI) 0.71 to 0.95). This reduction corresponds to an absolute survival benefit of 4% at two years, from 77% to 81%, and of 6% at five years, from 56% to 62%. There was significant heterogeneity between trials (P = 0.03; I² = 50%) largely due to the timing of chemotherapy (P = 0.005). The concomitant trials showed a better treatment effect than induction trials or adjuvant trials (HR = 0.60, 95%CI 0.48 to 0.76) versus HR = 0.99 (CI 0.80 to 1.21) and HR = 0.97 (CI 0.69 to 1.38)). The proportion of early deaths, i.e. within the three months after randomisation, was 1.6%, in the radio‐chemotherapy group and 1.2% in the radiotherapy alone group. The only excess treatment‐related deaths were observed in the radio‐chemotherapy group in induction chemotherapy trials. The survival curves can be found in Baujat 2006.

Effect of chemotherapy on event‐free survival

The number of events observed was 1044. There was a significant 24% reduction in the hazard ratio of tumour failure or death (P < 0.00001) for event‐free survival with chemotherapy (HR = 0.76, 95% CI 0.67 to 0.86). This reduction corresponds to an absolute event‐free survival benefit of 9% at two years, from 54% to 63%, and of 10% at five years, from 42% to 52%. There was no significant heterogeneity between trials (P = 0.14; I² = 32.5%) and no significant interaction between timings of chemotherapy (P = 0.09). HR = 0.82 (95% CI 0.68 to 0.97) in the induction trials, HR = 0.63 (95% CI 0.51 to 0.78) in the concomitant trials, and HR = 0.90 (95% CI 0.67 to 1.20) in the adjuvant trials.

Events were loco‐regional failure (46%), distant failure (38%), both loco‐regional and distant failure (5%), and death without failure (11%). Data on the type of failure were missing for one trial (Al‐Sarraf 1998). Chemotherapy lowered the risk of loco‐regional failure (P = 0.003, HR = 0.76, 95% CI 0.63 to 0.91) as well as that of the distant failure (P = 0.001, HR = 0.72, 95% CI 0.59 to 0.88). There was no significant interaction between the timing of chemotherapy and loco‐regional control (P = 0.89), nor between the timing of chemotherapy and distant control (P = 0.19). The event‐free survival curves can be found in Baujat 2006.

Interactions between treatment effect and trial characteristics

A significant beneficial effect of chemotherapy was observed on both event‐free survival and overall survival in the subset of six trials (730 patients; two induction, one concomitant plus adjuvant and three adjuvant trials) using cisplatin + 5‐FU: HR of tumour failure or death was 0.69 (95% CI 0.56 to 0.85) and HR of death, 0.74 (95% CI 0.59 to 0.93). A significant benefit was observed for event‐free survival, but not for overall survival in the other subset of five trials (1245 patients; two induction and three concomitant trials): HR of tumour failure or death, 0.80 (95% CI 0.69 to 0.93) and HR of death, 0.87 (95% CI 0.73 to 1.04). However, the difference between these treatment effects was not significant either for event‐free survival (P = 0.25) or for overall survival (P = 0.28).

Interactions between treatment effect and patient characteristics

Table 2 summarises the statistical analyses exploring the interactions between patient characteristics and the treatment effect. There was no significant interaction between the treatment effect and age, gender performance status, T stage or N stage. No significant interaction was found between T stage or N stage and the treatment effect within each subset of trials grouped according to the timing of chemotherapy. The only significant interaction was between the WHO histologic type and the effect of chemotherapy: chemotherapy was more efficient against WHO type 1 disease than against WHO type 2, 3 disease (P = 0.003 for overall survival and P < 0.0001 for event‐free survival).

2. Treatment effect on overall and event‐free survival according to patient charact.

Characteristics	n. patients RT+CT/RT	HR of death (95%CI)	P‐value	HR tum. failur/death	P‐value
Gender: Male	742/727	0.81 (0.69 to 0.95)		0.76 (0.66 to 0.87)
Gender: Female	248/258	0.85 (0.62 to 1.16)	0.81	0.74 (0.58 to 0.96)	0.89
Age < 40 or equal	326/285	0.85 (0.63 to 1.14)		0.67 (0.52 to 0.85)
41 to 50	308/327	0.77 (0.59 to 1.01)		0.80 (0.64 to 1.00)
> 50	356/373	0.86 (0.70 to 1.05)	0.85 (t for trend)	0.79 (0.66 to 0.95)	0.31 (t for trend)
Performance status: 0	380/368	0.89 (0.71 to 1.11)		0.78 (0.64 to 0.94)
1	342/340	0.71 (0.55 to 0.92)		0.66 (0.53 to 0.83)
2	17/21	1.55 (0.65 to 3.69)	0.73 (t for trend)	1.40 (0.65 to 3.02)	0.92 (t for trend)
T stage (AJCC97): T1	267/272	0.68 (0.51 to 0.90)		0.69 (0.54 to 0.87)
T2	350/363	0.83 (0.64 to 1.07)		0.82 (0.66 to 1.02)
T3/T4	373/350	0.90 (0.73 to 1.12)	0.12 (t for trend)	0.73 (0.60 to 0.88)	0.80 (t for trend)
N stage (AJCC97): N0	91/83	1.02 (0.61 to 1.69)		0.65 (0.42 to 1.00)
N1/N2	620/643	0.82 (0.68 to 0.99)		0.79 (0.68 to 0.93)
N3	242/219	0.68 (0.52 to 0.88)	0.24 (t for trend)	0.64 (0.51 to 0.81)	0.47 (t for trend)
WHO type 1	29/26	0.30 (0.15 to 0.59)		0.18 (0.09 to 0.36)
WHO type 2 to 3	958/959	0.85 (0.73 to 0.98)	0.003	0.78 (0.69 to 0.89)	< 0.0001
Total	990/985	0.82 (0.71 to 0.94)	0.006	0.76 (0.67 to 0.86)	< 0.0001

Sensitivity analyses

Sensitivity analyses were performed to check the robustness of the results (Table 3). After exclusion of patients with WHO type 1 disease, the overall result remained significantly in favour of chemotherapy (P = 0.03), and this exclusion diminished the heterogeneity between trials, which was no longer significant (P = 0.09). As 49 out of 55 patients with WHO type 1 disease were from the Al‐Sarraf 1998 trial, analyses were also performed after exclusion of this trial. The overall benefit of chemotherapy remained significant for event‐free survival (P = 0.002), but not for overall survival (P = 0.17). However, the treatment effect remained significant for the concomitant subset even though the hazard ratio increased from 0.60 (95% CI 0.48 to 0.76) to 0.71 (95% CI 0.53 to 0.94). Excluding a small trial, two trials with less than five years of follow up, and the two comparisons using a control group that received chemotherapy did not significantly modify overall results (Table 3).

3. Sensitivity analyses.

Trials included	N patients RT+CT/RT	OS. HR. 95%CI	OS. HR. p‐value	OS. heterogeneity I²	OS. Hetero. p‐value	EFS. HR. 95%CI	EFS. HR. p‐value	EFS. hetero. I²	EFS. Hetero. p‐value
All trials	990/985	0.82 (0.71 to 0.94)	0.006	50%	0.03	0.76 (0.67 to 0.86)	< 0.0001	32%	0.14
Without Al‐Sarraf 1998 (INT‐0099)	893/889	0.90 (0.77 to 1.05)	0.17	0%	0.37	0.82 (0.72 to 0.93)	0.002	0%	0.99
Without patients with WHO 1 carcinoma	958/959	0.85 (0.73 to 0.98)	0.03	38%	0.09	0.78 (0.69 to 0.89)	0.0001	0%	0.58
Without one small trial (Chan 1995 (PWH‐88))	953/945	0.81 (0.70 to 0.93)	0.003	51%	0.03	0.75 (0.66 to 0.85)	< 0.0001	36%	0.12
Without QMH‐95 combined arms (Kwong 2004d (QMH‐95conc+), Kwong 2004c (QMH‐95adj+))	876/875	0.84 (0.73 to 0.98)	0.02	53%	0.03	0.75 (0.66 to 0.85)	< 0.0001	43%	0.08
Without Chan 1995 (PWH‐88), QMH95: follow up < 5 years	729/725	0.80 (0.68 to 0.93)	0.004	58%	0.04	0.73 (0.64 to 0.84)	< 0.0001	32%	0.03

Discussion

Despite numerous trials investigating the effect of chemotherapy on nasopharyngeal carcinoma, to date there is no consensus about the magnitude of its benefit and the optimal protocol. An individual patient data meta‐analysis was therefore justified. The exhaustiveness principle of meta‐analysis was impossible to reach because of the difficulties encountered when trying to include Chinese trials. This is why the term "pooled analysis" could also be applied to the present study. The quality of the missing data remains unknown, while the quality of our data has been thoroughly checked. Very few data are missing from the trials included which comply with long‐term follow up. Data on 11 trials and 2793 patients were collected. The quality of the trials, especially concerning randomisation, was verified. Three trials totalling 1040 patients were excluded because they did not fulfil the eligibility criterion of unpredictable treatment assignment (Lin 2003; Ma 2001; Pan 2000). Our meta‐analysis demonstrates a small but significant treatment effect in terms of overall survival and event‐free survival. These results seem to be robust, as confirmed by our sensitivity analyses. It is noteworthy that chemotherapy also seems to be active in terms of loco‐regional control (HR = 0.76) and distant control (HR = 0.72).

A second meta‐analysis based on published results was reported recently (Langedijk 2004). Its results included two of the three trials that we excluded (Lin 2003; Ma 2001), but it did not include the results of the Kwong 2004a trial nor the overall survival data from the Chan 2002 trial. The individual patient data from an old trial (Rossi 1988) included in their meta‐analysis were lost and therefore were not included in our results. Literature‐based meta‐analyses tend to have limitations: no quality control of data, analyses are not based on the intention‐to‐treat principle in all the trials, analyses of the interaction between prognostic factors and treatment effects are not possible, and trials are not necessarily updated. These differences may explain why the effect of chemotherapy observed in our meta‐analysis in the concomitant group of trials is smaller than the effect observed in their meta‐analysis. However, in that study, the observed effect of chemotherapy on overall survival is close to our findings.

The treatment effect could be dependent on the timing of chemotherapy: no evidence of an overall survival benefit was observed with induction and adjuvant chemotherapy, unlike that evidenced with concomitant chemotherapy. A benefit for event‐free survival was however demonstrated in the subset of trials with induction chemotherapy. In this group, there was an excess of treatment‐related deaths in the chemo‐radiotherapy group. This may suggest that if toxicity was better managed, which is the case in the more recent trials, induction chemotherapy may play a role. The observed effect in the subset of trials with adjuvant chemotherapy could be attributable to the absence of an effect on distant control. However, the power of our study may be insufficient to demonstrate this hypothesis. The Al‐Sarraf 1998 trial, in which adjuvant chemotherapy was added to concomitant chemotherapy, was included in the concomitant chemotherapy group. However, the impact of adjuvant chemotherapy in this trial remains unclear, as there was poor compliance with adjuvant chemotherapy which was also the case in the Chan 1995 trial (adjuvant after induction chemotherapy): only 55% patients in these trials completed the adjuvant chemotherapy protocol. In addition, when adjuvant chemotherapy was used alone, the trials failed to demonstrate a positive impact either in terms of overall survival or event‐free survival. In the present meta‐analysis, the Al‐Sarraf 1998 trial was the only trial to demonstrate a significant overall survival benefit imputable to chemotherapy. Some controversy arose about the relevance of this trial because 49 of the 193 patients had WHO type 1 carcinoma, and survival in the control group was lower than usual (Chua 2000). Furthermore, trial accrual was stopped early due to the highly significant survival benefit evidenced at the time of the first planned interim analysis. Stopping the trial early may have artificially led to an overestimation of the treatment effect. Two trials should clarify this point: the SQNP01 trial conducted by the Nasopharynx Cancer work group in Singapore (Wee 2005a) confirms the results of the Al‐Sarraf 1998 trial: a significant benefit in favour of chemotherapy is observed (HR = 0.51; 95% CI 0.31 to 0.81; P = 0.006). The compliance with adjuvant chemotherapy is also low in this trial: only 57% of the patients completed the adjuvant chemotherapy protocol. The preliminary results of the trial conducted by the Hong Kong Nasopharyngeal Cancer Study group and the Princess Margaret Hospital in Canada (Lee 2005a) do not demonstrate any effect of this protocol on overall survival. A longer follow up is needed for this trial. However, the result of our meta‐analysis cannot simply be attributable to the Al‐Sarraf 1998 trial, because the event‐free survival benefit for the whole group of trials and the overall survival benefit in the concomitant group afforded by chemotherapy remained significant after exclusion of this trial. In our meta‐analysis, only 55 patients had WHO type 1 carcinoma, and they account for part of the heterogeneity of the treatment effects. This heterogeneity may have stemmed from a significantly more pronounced treatment effect of chemotherapy in this subgroup of patients. As most of these patients were from the Al‐Sarraf 1998 trial where a strong treatment effect was observed, a bias may have been introduced, but this is probably not the case, because the treatment effect in this trial was also more pronounced among patients with WHO type 1 carcinoma. However, the difference in the treatment effect between patients with WHO type 1 carcinoma and those with WHO type 2 and 3 carcinoma was not significantly different. No other significant interaction between the effect of chemotherapy and patient characteristics could be found.

Authors' conclusions

Implications for practice.

The adjunction of chemotherapy to standard radiotherapy provides a small but significant survival benefit in patients with nasopharyngeal carcinoma. This benefit is essentially observed when chemotherapy is administered concomitantly with radiotherapy. The role of induction chemotherapy and adjuvant chemotherapy given alone or added to concomitant or induction chemotherapy is more questionable.

Implications for research.

Due to the heterogeneity of the chemotherapy protocols, this meta‐analysis does not allow conclusions on the superiority of a chemotherapy regimen. Furthermore, the different timings used to deliver different drugs can interfere with the results. Induction chemotherapy trials showed a benefit in event‐free survival and the lack of effect on overall survival may find an explanation in the excess of toxic deaths. Following the positive results of taxane‐based induction chemotherapy protocols in head and neck squamous cell carcinoma, it may be useful to design trials investigating induction chemotherapy plus concomitant chemotherapy versus concomitant chemotherapy alone in nasopharyngeal cancer.

Feedback

Feedback received, 16 February 2015

Summary

This review contains a trial with multiple treatment arms (Kwong, 2004). It appears that the groups of patients from Kwong (2004) have been included more than in the same meta‐analysis, leading to unit of analysis issues (see Cochrane handbook section 9.3.9).

I agree with the conflict of interest statement below:

I certify that I have no affiliations with or involvement in any organization or entity with a financial interest in the subject matter of my feedback.

Reply

MAC‐NPC was an individual patient‐based meta‐analysis, meaning that the authors of the meta‐analysis obtained the patients' data from all trials included and performed intention‐to treat analyses, after a rigorous verification process.

The Kwong trial is a 2 x 2 design trial, thus allowing four comparisons: 'RT versus RT + concomitant CT'; 'RT versus RT + adjuvant CT'; 'RT + concomitant CT versus RT + concomitant CT + adjuvant CT'; and 'RT + adjuvant CT versus RT + adjuvant + concomitant CT'. The 222 patients in this trial, out of the 1753 patients in the meta‐analysis were thus counted twice (+13%). This is commonly performed in other meta‐analyses (see e.g. MACH‐NC, MARCH). Exclusion of the two following comparisons, 'RT + concomitant CT versus RT + concomitant CT + adjuvant CT' and 'RT+ adjuvant CT versus RT + adjuvant + concomitant CT', did not modify its conclusions. This meta‐analysis has been updated and will be published very soon in Lancet Oncology.

Bertrand Baujat, 25 March 2015.

Contributors

Comment from: Nathan Bromham, Researcher, National Collaborating Centre for Cancer, nathan.bromham@wales.nhs.uk.

Reply from: Dr Bertrand Baujat, Service ORL‐CCF, Hôpital Tenon, Paris, France.

What's new

Date	Event	Description
27 April 2015	Feedback has been incorporated	Author reply to feedback added.

History

Protocol first published: Issue 3, 2003
 Review first published: Issue 4, 2006

Date	Event	Description
24 March 2015	Amended	Statement about unrestricted pharmaceutical company grants moved from 'Acknowledgements' to Declarations of interest.
24 March 2015	Feedback has been incorporated	Feedback received.
1 May 2009	Amended	Published note added regarding the future updating of this review
5 November 2008	Amended	Converted to new review format.

Notes

May 2009: updating of this review

This is a review of individual patient data.  Whilst standard Cochrane Reviews are updated every two years, we believe that the updating of this review is premature because:

• induction chemotherapy is a major issue since the onset of TPF (taxane, platinum, 5FU) protocols. We know of two ongoing trials which are testing TPF for nasopharyngeal carcinoma and think it would be pertinent to defer the update of this review until they publish at least their preliminary results;

• two newly published trials confirm the results of the previously published trials and no different conclusions would be drawn from their inclusion. This systematic review was not underpowered, as concomitant chemotherapy was shown to provide a significant benefit in terms of survival.

Data and analyses

Comparison 1. Radiotherapy versus radiotherapy + chemotherapy.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Effect of chemotherapy on overall survival, hazard ratio of death by timing of chemotherapy	11	1975	Peto Odds Ratio (95% CI)	0.82 [0.71, 0.95]
1.1 Induction +/‐ adjuvant chemotherapy	4	830	Peto Odds Ratio (95% CI)	0.99 [0.80, 1.21]
1.2 Concomitant +/‐ adjuvant chemotherapy	4	765	Peto Odds Ratio (95% CI)	0.60 [0.48, 0.76]
1.3 Adjuvant chemotherapy	3	380	Peto Odds Ratio (95% CI)	0.97 [0.68, 1.38]
2 Effect of chemotherapy on event‐free survival, hazard ratio of tumour failure or death by timing of chemo.	11	1975	Peto Odds Ratio (95% CI)	0.76 [0.67, 0.86]
2.1 Induction +/‐ adjuvant chemotherapy	4	830	Peto Odds Ratio (95% CI)	0.82 [0.68, 0.97]
2.2 Concomitant +/‐adjuvant chemotherapy	4	765	Peto Odds Ratio (95% CI)	0.63 [0.51, 0.78]
2.3 Adjuvant chemotherapy	3	380	Peto Odds Ratio (95% CI)	0.90 [0.67, 1.20]
3 Effect of chemotherapy on loco‐regional control, HR of loco‐regional failure by timing of chemotherapy	10	1782	Peto Odds Ratio (95% CI)	0.76 [0.63, 0.91]
3.1 Induction +/‐ adjuvant chemotherapy	4	830	Peto Odds Ratio (95% CI)	0.76 [0.60, 0.97]
3.2 Concomitant +/‐ adjuvant chemotherapy	3	572	Peto Odds Ratio (95% CI)	0.81 [0.55, 1.18]
3.3 Adjuvant chemotherapy	3	380	Peto Odds Ratio (95% CI)	0.71 [0.48, 1.04]
4 Effect of chemotherapy on distant control, hazard ratio of distant failure by timing of chemotherapy	10	1782	Peto Odds Ratio (95% CI)	0.72 [0.59, 0.87]
4.1 Induction +/‐ adjuvant chemotherapy	4	830	Peto Odds Ratio (95% CI)	0.65 [0.49, 0.86]
4.2 Concomitant +/‐ adjuvant chemotherapy	3	572	Peto Odds Ratio (95% CI)	0.69 [0.49, 0.97]
4.3 Adjuvant chemotherapy	3	380	Peto Odds Ratio (95% CI)	1.11 [0.66, 1.85]

1.1. Analysis.

Comparison 1 Radiotherapy versus radiotherapy + chemotherapy, Outcome 1 Effect of chemotherapy on overall survival, hazard ratio of death by timing of chemotherapy.

1.2. Analysis.

Comparison 1 Radiotherapy versus radiotherapy + chemotherapy, Outcome 2 Effect of chemotherapy on event‐free survival, hazard ratio of tumour failure or death by timing of chemo..

1.3. Analysis.

Comparison 1 Radiotherapy versus radiotherapy + chemotherapy, Outcome 3 Effect of chemotherapy on loco‐regional control, HR of loco‐regional failure by timing of chemotherapy.

1.4. Analysis.

Comparison 1 Radiotherapy versus radiotherapy + chemotherapy, Outcome 4 Effect of chemotherapy on distant control, hazard ratio of distant failure by timing of chemotherapy.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Al‐Sarraf 1998.

Methods	Inclusion period: 1989 to 1995. Median follow up: 110 months.
Participants	193 patients. Stages III to IV (AJCC < 1997). Histology WHO 1 to 3.
Interventions	Radiotherapy: Tumor: 70 Gray/7 weeks. N‐ 50 Gray, N+ 66 to 70 Gray +/‐ concomitant Cisplatin: 100 mg/m² x 3 and adjuvant Cisplatin: 80 mg/m² x 3 cycles and Fluorouracil: 4000 mg/m² x 3 cycles, continuous infusion.
Outcomes	Overall survival, event‐free survival.
Notes	24% of type 1 histology, concomitant cisplatin every 3 weeks, adjuvant cisplatin + 5‐fluorouracil every 4 weeks.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	Low risk	A ‐ Adequate

Chan 1995.

Methods	Inclusion period: 1988 to 1991. Median follow up: 35 months.
Participants	77 patients. Stage III to IV (Ho classification). Histology WHO 3.
Interventions	Radiotherapy: Tumor: 66 Gray/6.5 weeks. N‐ 58 Gray, N+ 65.5 Gray +/‐ induction and adjuvant chemotherapy: Cisplatin: 100 mg/m² x (induction 2 cycles and adjuvant 4 cycles) and Fluorouracil: 3000 mg/m² x (induction 2 cycles and adjuvant 4 cycles), continuous infusion.
Outcomes	Overall survival, event‐free survival.
Notes	2 cycles of induction and 4 cycles of adjuvant chemotherapy. Radiotherapy : nasopharynx, equivalent of 66 Gy with conventional fractionation + 20 Gy boost if parapharyngeal disease + 18 to 24 Gy using 192Ir if residual disease 4 weeks after radiotherapy; neck: 58 Gy for lower neck, 66 Gy for upper neck.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	Low risk	A ‐ Adequate

Chan 2002.

Methods	Inclusion period: 1994 to 1999. Median follow up: 67 months.
Participants	350 patients. Stages II to IV (AJCC 1997). Histology WHO 1 to 3.
Interventions	Radiotherapy: Tumor: 66 Gray/6.5 weeks. N‐ 58 Gray, N+ 65.5 Gray +/‐ concomitant Cisplatin: 40 mg/m² , weekly.
Outcomes	Overall survival, event‐free survival.
Notes	10 or 20 Gy (depending on the centre) boost if parapharyngeal disease, 21 to 24 Gy using 192Ir if residual local disease after radiotherapy, 7.5 Gy boost if residual nodal disease, radical neck dissection if proven residual neck nodes.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	Low risk	A ‐ Adequate

Chi 2002.

Methods	Inclusion period: 1994 to 1999. Median follow up: 72 months.
Participants	158 patients. Stage IV (AJCC < 1997). Histology WHO 1 to 3.
Interventions	Radiotherapy: Tumor: 70 to 72 Gray/7 to 8 weeks. N‐ 50 Gray lower neck +/‐ Adjuvant Cisplatin: 20 mg/m² x 9 weekly, Fluorouracil: 2200 mg/m² x 9 weekly, Leucovorin acid: 120 mg/m² x 9 weekly.
Outcomes	Overall survival, event‐free survival.
Notes	All drugs provided by continuous injections.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	Low risk	A ‐ Adequate

Chua 1998.

Methods	Inclusion period: 1989 to 1993. Median follow up: 65 months.
Participants	334 patients. Stages II to IV (AJCC < 1997). Histology WHO 2 to 3.
Interventions	Radiotherapy: Tumor 66 to 74 Gray/6.5 to 7.5 weeks. N‐ 60 to 66 Gray, N+ 66 to 76 Gray. +/‐ Induction chemotherapy: Cisplatin: 60 mg/m² x 2 to 3; Epirubicin: 110 mg/m² x 2 to 3.
Outcomes	Overall survival, event‐free survival.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	Low risk	A ‐ Adequate

Cvitkovic 1996.

Methods	Inclusion period: 1989 to 1993. Median follow up: 84 months.
Participants	339 patients. Stages III to IV (AJCC < 1997). Histology WHO 1 to 3.
Interventions	Radiotherapy: Tumor 65 to 70 Gray/6.5 to 7.5 weeks. N‐ 50 Gray, N+ 65 Gray. +/‐ Induction chemotherapy: Bleomycin: 15 mg/m² x 3; Bleomycin: 60 mg/m² x 3, continuous injection; Epirubicin: 70 mg/m² x 3; Cisplatin: 100 mg/m² x 3.
Outcomes	Overall survival, event‐free survival.
Notes	110 patients treated with conventional radiotherapy and 176 patients with hypofractionated radiotherapy, 2.5 Gy x 3/weeks followed by 3.5 Gy x 3/week.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	Low risk	A ‐ Adequate

Hareyama 2002.

Methods	Inclusion period: 1991 to 1998. Median follow up: 74 months.
Participants	80 patients. Stages I to IV (AJCC < 1997). Histology WHO 1 to 3.
Interventions	Radiotherapy: Tumor: 66 to 68 Gray/6.5 to 7 weeks. N‐ 50 Gray, N+ 66 to 68 Gray +/‐ induction chemotherapy: Cisplatin: 80 mg/m² x 2 cycles and Fluorouracil: 3200 mg/m² x 2 cycles, continuous infusion.
Outcomes	Overall survival, event‐free survival.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	Low risk	A ‐ Adequate

Kwong 2004a.

Methods	Inclusion period: 1995 to 2000. Median follow up: 57 months
Participants	222 patients. Stages II to IV (AJCC 1997). Histology WHO 1 to 3.
Interventions	Radiotherapy: Tumor: 62.5 to 68 Gray/7 weeks. N 62.5 to 66 Gray/7weeks +/‐ boost 10 Gray +/‐ concomitant UFT 600 mg daily po / Adjuvant Cisplatin: 100 mg/m² x 3, Fluorouracil: 3000 mg/m² x 3, Vincristine: 2mg x 3, Bleomycin: 30 mg x 3, Methotrexate: 150 mg/m² x 3
Outcomes	Overall survival, Event‐free survival
Notes	2 X 2 design, concomitant chemotherapy versus none, adjuvant chemotherapy versus none, concomitant+adjuvant chemotherapy versus adjuvant chemotherapy, concomitant +adjuvant chemotherapy versus concomitant chemotherapy ; for adjuvant chemotherapy, alternating cycles of cisplatin + 5‐fluorouracil and vincristine + bleomycin + methotrexate. +/‐ 10 Gy additional boost in case of parapharyngeal space involvement and/or palpable residual nodes.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	Low risk	A ‐ Adequate

Kwong 2004b.

Methods	See Kwong 2004a
Participants
Interventions
Outcomes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	Low risk	A ‐ Adequate

Kwong 2004c.

Methods	See Kwong 2004a
Participants
Interventions
Outcomes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	Low risk	A ‐ Adequate

Kwong 2004d.

Methods	See Kwong 2004a
Participants
Interventions
Outcomes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	Low risk	A ‐ Adequate

Kwong 2004a (QMH‐95 conc), Kwong 2004b (QMH‐95adj), Kwong 2004c (QMH‐95adj+), Kwong 2004d (QMH‐95conc+): 4 comparisons issued from QMH‐95 2x2 trial.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Lin 2003	Allocation: not properly randomised.
Ma 2001	Allocation: not properly randomised.
Pan 2000	Allocation: not properly randomised.
Rossi 1988	Individual data lost by the institution.

Characteristics of studies awaiting assessment [ordered by study ID]

Chen 2008.

Methods	Randomized phase III trial
Participants	316 (158 per group) locoregionally advanced nasopharyngeal carcinomas
Interventions	Radiotherapy (70 Grays in 7 weeks) plus concurrent cisplatin (40 mg/m² weekly) followed by adjuvant cisplatin (80 mg/m² on day 1) + 5 Fluoro‐Uracil (800 mg/m² on days 1 to 5) every 4 weeks for 3 cycles versus same radiotherapy
Outcomes	Significant benefit in favour of chemotherapy arm on: 2‐year overall survival rate (89.8% versus 79.7%, P = 0.003) 2‐year failure‐free survival rate (84.6% versus 72.5%, P = 0.001) 2‐year distant failure‐free survival rate (86.5% versus 78.7% P = 0.024) 2‐year locoregional failure‐free survival rate (98% versus 91.9%, P = 0.007)
Notes	Notes: more toxicity and lower compliance in the chemotherapy arm

Hui 2009.

Methods	Randomized phase II trial
Participants	65 (34 CT + CRT versus 31 CRT) stage III to IVb nasopharyngeal carcinomas
Interventions	Induction docetaxel (75 mg/m²) + cisplatin (75 mg/m²) every 3 weeks for 2 cycles followed by radiotherapy (conventional or IMRT) + concurrent cisplatin (40 mg/m²) weekly versus same radiotherapy + same concurrent cisplatin
Outcomes	Significant benefit on overall survival in favour of induction chemotherapy arm: 3‐year overall survival: 94.1% versus 67.7%, Hazard Ratio = 0.24 (95% CI 0.078 to 0.73), P = 0.012 3‐year progression‐free survival 88.2% versus 59.5%, Hazard Ratio = 0.49 (95% CI 0.20 to 1.19), P = 0.12
Notes	Notes: induction CT followed by concurrent RT + CT was well tolerated

Lee 2005.

Methods
Participants	335
Interventions	RT versus RT + Concomitant Cisplatin + Adjuvant Cisplatin + 5FU versus RT + Induction CT
Outcomes	No significant benefit on overall survival
Notes	See footnotes

Wee 2005.

Methods
Participants	221
Interventions	RT versus RT + Concomitant Cisplatin + Adjuvant Cisplatin + 5FU
Outcomes	Significant benefit from CT
Notes	See footnotes

Characteristics of ongoing studies [ordered by study ID]

Lee 2006.

Trial name or title	NPC99‐02
Methods
Participants	189
Interventions	RT versus accelerated RT versus RT + Concomitant Cisplatin + Adjuvant Cisplatin + 5FU versus accelerated RT + Concomitant Cisplatin + Adjuvant Cisplatin + 5FU
Outcomes	Preliminary results in favour of accelerated RT + chemotherapy
Starting date	1999
Contact information	Lee AW, Hong Kong Nasopharyngeal Cancer Study Group and the Princess Margaret Hospital in Canada
Notes	See footnotes

VUMCA II.

Trial name or title	VUMCA II
Methods
Participants	509
Interventions	Induction CT+ RT versus Induction CT + RT + concomitant CT
Outcomes	Not ready for analysis
Starting date	1996
Contact information	Institut Gustave‐Roussy Dr E. Benhamou
Notes

Zhang 2005.

Trial name or title	Sun Yat‐Sen University
Methods
Participants	77
Interventions	RT versus RT + concomitant oxaliplatin
Outcomes	Significant benefit from CT
Starting date	2001
Contact information	Cancer Centre of Sun Yat Sen University
Notes	See footnotes

Recently published as a full paper:
 ‐ Lee AW, Lau W H, Tung S Y, Chua D, Chappell R, Xu L, Siu L et al. Preliminary results of a randomized study on therapeutic gain by concurrent chemotherapy for regionally advanced Nasopharyngeal Carcinoma: NPC‐9901 trial by the Hong‐Kong Nasopharyngeal Cancer Study Group. J Clin Oncol 2005; 23: 6966‐75. 
 ‐ Wee J, Tan EH, Tia BC, Wong H B, Leong S S, Tan T, Chua E T, Yang E, Lee K M,.Fong KW et al. Phase III randomized trial of radiotherapy versus concurrent chemo‐radiotherapy followed by adjuvant chemotherapy in patients with AJCC/UICC (1997) stage 3 and 4 nasopharyngeal cancer of the endemic variety. J Clin Oncol 2005; 23: 6730‐38.
 ‐ Zhang L, Zhao C, Peng PJ, Lu LX, Han F, Wu SX. Phase III study comparing standard radiotherapy with or without weekly oxaliplatin in treatment of locoregionally advanced nasopharyngeal carcinoma: preliminary results. J Clin Oncol 2005; 23: 8461‐68.
 ‐ Lee AW, Tung SY, Chan AT, Chappell R, Fu YT, Lu TX, Tan T, Chua DT, O'Sullivan B, Xu SL, Pang ES, Sze WM, Leung TW, Kwan WH, Chan PT, Liu XF, Tan EH, Sham JS, Siu L, Lau WH. Preliminary results of a randomized study (NPC‐9902 Trial) on therapeutic gain by concurrent chemotherapy and/or accelerated fractionation for locally advanced nasopharyngeal carcinoma. Int J Radiat Oncol Biol Phys. 2006 ;66:142‐51.

Contributions of authors

B Baujat supervised the data analysis and wrote the paper. H Audry did the data collection and the statistical analysis. J Bourhis and JP Pignon had the original idea, designed the study and revised the paper. JP Pignon also supervised data collection and statistical analyses. The investigators provided the data, answered queries sent to secretariat and validated the re‐analysis of trials. The other authors reviewed and commented on the preliminary results and the manuscript. The members of the Steering Committee also reviewed the protocol and the data from trials with eligibility problems. K Thephamongkhol contributed to data collection and revised the paper.

Sources of support

Internal sources

• Institut Gustave‐Roussy, France.

External sources

• Aventis, France.

• Sanofi‐Synthelabo, France.

• Schering‐Plough, France.

Declarations of interest

Aventis had a general agreement with Institut Gustave Roussy to support meta‐analyses on lung and head and neck cancers. Sanofi supported part of the investigator meeting. Schering‐Plough (a company not directly interested in the project) provided a travelling expenses grant for the first author to attend the ASTRO meeting to present the MAC‐NPC project. The grant was awarded after the acceptance of the abstract. The study sponsors played no role in the study design, data collection, data analysis, data interpretation or writing of the report.

The authors declare that they have no conflict of interest.

Edited (no change to conclusions), comment added to review