Hyperfractionated or accelerated radiotherapy for head and neck cancer

Abstract

Background

Several trials have studied the role of altered fractionation radiotherapy in head and neck squamous cell carcinoma, but the effect of such treatment on survival is not clear.

Objectives

The aim of this individual patient data (IPD) meta‐analysis was to assess whether this type of radiotherapy could improve survival.

Search methods

We searched the Cochrane Ear, Nose and Throat Disorders Group Trials Register; CENTRAL (2010, Issue 3); PubMed; EMBASE; CINAHL; Web of Science; BIOSIS Previews; Cambridge Scientific Abstracts; ISRCTN and additional sources for published and unpublished trials. The date of the most recent search was 8 August 2010.

Selection criteria

We identified randomised trials comparing conventional radiotherapy with hyperfractionated or accelerated radiotherapy, or both, in patients with non‐metastatic head and neck squamous cell carcinomas and grouped trials into three pre‐specified treatment categories: hyperfractionated, accelerated and accelerated with total dose reduction. Trials were eligible if they began recruitment after 1969 and ended before 1998.

Data collection and analysis

We obtained updated individual patient data. Overall survival was the main outcome measure. The secondary outcome measures were local or regional control rates (or both), distant control rates and cause‐specific mortality.

Main results

We included 15 trials with 6515 patients. The median follow up was six years. Tumour sites were mostly oropharynx and larynx; 5221 (74%) patients had stage III‐IV disease (UICC 2002). There was a significant survival benefit with altered fractionation radiotherapy, corresponding to an absolute benefit of 3.4% at five years (hazard ratio (HR) 0.92, 95% CI 0.86 to 0.97; P = 0.003). The benefit was significantly higher with hyperfractionated radiotherapy (8% at five years) than with accelerated radiotherapy (2% with accelerated fractionation without total dose reduction and 1.7% with total dose reduction at five years, P = 0.02). There was a benefit in locoregional control in favour of altered fractionation versus conventional radiotherapy (6.4% at five years; P < 0.0001), which was particularly efficient in reducing local failure, whereas the benefit on nodal control was less pronounced. The benefit was significantly higher in the youngest patients (under 50 year old) (HR 0.78, 95% CI 0.65 to 0.94), 0.95 (95% CI 0.83 to 1.09) for 51 to 60 year olds, 0.92 (95% CI 0.81 to 1.06) for 61 to 70 year olds, and 1.08 (95% CI 0.89 to 1.30) for those over 70 years old; test for trends P = 0.007).

Authors' conclusions

Altered fractionation radiotherapy improves survival in patients with head and neck squamous cell carcinoma. Comparison of the different types of altered radiotherapy suggests that hyperfractionation provides the greatest benefit. An update of this IPD meta‐analysis (MARCH 2), which will increase the power of this analysis and allow for other comparisons, is currently in progress.

Plain language summary

Hyperfractionated or accelerated radiotherapy for head and neck cancer

Radiotherapy is often used to treat head and neck cancers. The dosage of radiation is measured in Gray (Gy). When radiotherapy is given alone, the most commonly used schedule is 2 Gy in a single fraction per day, five days a week, for seven weeks. However, alternative radiotherapy regimens to reduce the total treatment time for head and neck cancers have been assessed. 'Acceleration' of the treatment (delivering the same total dose in a shorter time) should reduce the regrowth of the tumour between sessions, resulting in improved local control of the disease. In 'hyperfractionated' regimens, two to three fractions are delivered each day, with a reduced dose per fraction equal to 1.1 to 1.2 Gy. The reduction of the dose per fraction may reduce the risk of late toxicity, despite an increased total dose. Acceleration and hyperfractionation can be combined, in particular for regimens in which overall treatment time is reduced.

This Cochrane Review is an individual patient data based meta‐analysis and the aim was to assess whether this type of radiotherapy could improve survival. We identified randomised trials comparing conventional radiotherapy with hyperfractionated or accelerated radiotherapy, or both, in patients with non‐metastatic head and neck cancers and grouped trials into three pre‐specified categories: hyperfractionated, accelerated without total dose reduction and accelerated with total dose reduction. The results of this meta‐analysis suggest that altered fractionation radiotherapy improves survival in patients with head and neck cancer. Comparison of the different types of altered fractionation radiotherapy suggests that hyperfractionation provides the greatest benefit.

Individual patient data meta‐analysis is a long process and this review included all eligible trials which had completed recruiting patients by 1998. A major update of the analysis, including data from more recent trials, is currently underway.

Background

Head and neck squamous cell carcinomas are frequent tumours, with more than 550,000 new cases of oral cavity, oropharynx, hypopharynx and larynx cancer worldwide every year (Ferlay 2004). About 40% of patients have locally advanced disease at diagnosis.

Surgery, radiation therapy, or both, have been used for decades to achieve locoregional control. The most commonly used schedule, when radiotherapy is given alone, is 2 Gray (Gy) in a single fraction per day, five days a week, for seven weeks. Despite these treatments, the prognosis of patients with head and neck squamous cell carcinomas with locally advanced disease remains poor, with five‐year survival rates of 30% to 35% (Pignon 2000a).

In the past decade, new radiotherapy regimens for the treatment of head and neck squamous cell carcinomas have been assessed. These regimens were designed to increase the dose‐intensity by delivering a higher total dose in the same time (EORTC 22791 1992; PMHToronto 2007; RIO 1991; RTOG 9003HF 2000), the same total dose in five to six weeks instead of seven weeks (BCCA 9113 1997; CAIR 2000; DAHANCA 2003; EORTC 22851 1997; KBN PO 79 2002; Oro 9301 2003; RTOG 9003HF 2000; RTOG 9003B 2000; RTOG 9003S 2000) or a smaller total dose given in three to four weeks (CHART 1997; GORTEC 9402 2006; RTOG 7913 1987; TROG 9101 2001; Vienna 2000). Reducing the total treatment time, i.e. accelerating the treatment, should reduce the repopulation of tumour cells between sessions, resulting in improved locoregional control. In hyperfractionated regimens, two to three fractions are delivered each day, with a reduced dose per fraction equal to 1.1 to 1.2 Gy. The reduction of the dose per fraction might reduce the risk of late toxicity, despite an increased total dose.

Acceleration and hyperfractionation can be combined, in particular for regimens in which overall treatment time is reduced. In some randomised trials, altered fractionation radiotherapy has proved to be of benefit in locoregional control (CAIR 2000; DAHANCA 2003; EORTC 22791 1992; GORTEC 9402 2006; PMHToronto 2007; RIO 1991; RTOG 9003HF 2000) although no benefit in survival was generally detected. The use of altered fractionation radiotherapy is associated with some increase in toxicity, mostly due to mucositis (BCCA 9113 1997; DAHANCA 2003; EORTC 22851 1997; GORTEC 9402 2006; RTOG 9003HF 2000; RTOG 9003B 2000; RTOG 9003S 2000) and can add some practical constraints in radiotherapy departments, for example treatment two to three times a day or at a weekend, that need to be balanced by substantial benefit (BCCA 9113 1997; CAIR 2000; CHART 1997; DAHANCA 2003; EORTC 22791 1992; EORTC 22851 1997; GORTEC 9402 2006; KBN PO 79 2002; PMHToronto 2007; RIO 1991; RTOG 9003HF 2000; RTOG 9003B 2000; RTOG 9003S 2000; RTOG 7913 1987; TROG 9101 2001; Vienna 2000).

A meta‐analysis of updated individual patient data is the most reliable way to assess whether altered fractionation radiotherapy could affect survival. This meta‐analysis was undertaken by the MARCH (Meta‐Analysis of Radiotherapy in Carcinomas of Head and neck) Collaborative Group. A major update of this analysis (MARCH 2) is in progress.

Objectives

To study the effect of altered fractionation radiotherapy versus conventional radiotherapy on overall survival rates. We also planned a comparison between the effects of the three types of altered fractionation radiotherapy.

Methods

Criteria for considering studies for this review

Types of studies

Trials had to be randomised in a way that ensured that investigators decided whether the patient was eligible without foreknowledge of the assigned treatment. Trials were eligible if recruitment began after 1969 and ended before 1998.

Types of participants

Trials including previously untreated patients (those who had not received prior radiotherapy or prior chemotherapy), with non‐metastatic head and neck squamous cell carcinomas of the oral cavity, oropharynx, hypopharynx or larynx, were eligible. We excluded randomised trials including mainly or exclusively nasopharyngeal carcinomas as their epidemiology and response to radio‐ and chemotherapy is different from that of other head and neck cancers.

Types of interventions

Trials that compared conventional radiotherapy with accelerated or hyperfractionated radiotherapy, or both, in previously untreated patients with non‐metastatic head and neck squamous cell carcinomas were eligible. Trials had to be unconfounded: they should differ only on radiotherapy modalities. Definition of conventional dose was based on expert opinion and chosen during the investigator meeting before having seen the results: trials that used doses per fraction higher than 2.5 Gy were not eligible. Radiotherapy had to be with curative intent. We excluded trials without a conventional radiotherapy arm. We defined conventional curative radiotherapy as radiotherapy equivalent to 66 to 70 Gy, in 2 Gy fractions, for five days a week. We also excluded trials of postoperative radiotherapy from this meta‐analysis because of their small number and their different total dose.

The Collaborative Group's Steering Committee prespecified three groups of trials with different modifications of fractionation that correspond to three distinct biological questions. The first group (hyperfractionation) examined the effect of a higher total dose in the same overall time than in the reference arm (EORTC 22791 1992; PMHToronto 2007; RIO 1991; RTOG 9003HF 2000). The second group (accelerated group) represented a pure test of the effect of accelerating radiotherapy, while keeping the total dose the same (BCCA 9113 1997; CAIR 2000; DAHANCA 2003; EORTC 22851 1997; KBN PO 79 2002; Oro 9301 2003; RTOG 9003B 2000; RTOG 9003S 2000). Finally, the third group (accelerated with reduced dose group) tested the effect of accelerating radiotherapy, but with reduced total dose (CHART 1997; GORTEC 9402 2006; RTOG 7913 1987; TROG 9101 2001; Vienna 2000).

Types of outcome measures

Overall survival was the main outcome measure and was defined as the time from randomisation to the last follow up or death, whatever the cause.

The secondary outcome measures were local or regional control rates (or both), distant control rates and cause‐specific mortality.

Search methods for identification of studies

We conducted systematic searches for randomised controlled trials. There were no language, publication year or publication status restrictions. The date of the last search was 8 August 2010.

Electronic searches

We searched the following databases from their inception for published, unpublished and ongoing trials: the Cochrane Ear, Nose and Throat Disorders Group Trials Register; the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2010, Issue 3); PubMed; EMBASE; CINAHL; LILACS; KoreaMed; IndMed; PakMediNet; CAB Abstracts; Web of Science; BIOSIS Previews; CNKI; ISRCTN; ClinicalTrials.gov; ICTRP and Google.

We modelled subject strategies for databases on the search strategy designed for CENTRAL. Where appropriate, we combined subject strategies with adaptations of the highly sensitive search strategy designed by the Cochrane Collaboration for identifying randomised controlled trials and controlled clinical trials (as described in The Cochrane Handbook for Systematic Reviews of Interventions Version 5.0.2, Box 6.4.b. (Handbook 2009)). Search strategies for major databases including CENTRAL are provided in Appendix 1.

Searching other resources

We scanned the reference lists of identified publications for additional trials and contacted trial authors where necessary. In addition, we searched PubMed, TRIPdatabase, NHS Evidence ‐ ENT & Audiology and Google to retrieve existing systematic reviews relevant to this systematic review, so that we could scan their reference lists for additional trials. We searched for conference abstracts using the Cochrane Ear, Nose and Throat Disorders Group Trials Register and the websites for ASCO, ESMO, ASTRO, ECCO, ESTRO and PDQ. We also asked experts and trialists who took part in the meta‐analysis to identify trials.

Data collection and analysis

Both the Institute Gustave Roussy and the European Organisation for Research and Treatment of Cancer (EORTC) meta‐analysis unit performed data extraction. The data requested for all patients were age, sex, tumour site, T and N classification, stage, histology, performance status, allocated treatment and date of randomisation. The date and types of the first tumour failure (local, regional or distant) and the date of second primary cancer were also recorded. We requested updated survival status and date of last follow up from the trialists. We obtained data for patients excluded from the analysis after randomisation whenever possible. We did not gather data for morbidity and toxicity because this information was not available in a common format.

We used standard checks to identify missing data and assess data validity and consistency. We verified the amount of missing data and checked the order of dates. To assess randomisation integrity we checked patterns of treatment allocation and balance of baseline characteristics by treatment group. We checked follow up of surviving patients to ensure that it was balanced by treatment group and was up to date. We compared all data with the trial's protocol and published reports. We resolved any queries and each trial investigator or statistician verified the final database. We verified ranges and extremes with the trialists.

We also assessed study quality using the Cochrane 'Risk of bias' approach, which involved assessment of six domains: sequence generation; allocation concealment; blinding; incomplete outcome data; selective outcome reporting and other sources of bias. Assessment involved describing each of these domains as reported in the trial and then assigning a judgement about the adequacy of each entry. This meant answering a pre‐specified question whereby a judgement of 'Yes' indicated low risk of bias, 'No' indicated high risk of bias, and 'Unclear' indicated unclear or unknown risk of bias.

We analysed each trial individually and sent the survival analyses to the trialists for validation. We computed median follow up by the reverse Kaplan‐Meier method (Schemper 1996). We stratified survival analyses by trials and used the log‐rank observed minus expected numbers of deaths (O‐E) and their variances to calculate individual hazard ratios (HR) and overall HR with a fixed‐effect model. The weight of each trial in the pooled analysis was proportional to the variance of O‐E, which is roughly equal to a quarter of the number of deaths (Yusuf 1985). To eliminate the potential bias of an incorrect determination of the cause of death after recurrence, the log‐rank analysis of deaths from non‐head and neck cancer cause covered only the period before recurrence (i.e. data were censored at the first recurrence), as Peto and colleagues proposed (EBCT‐CG 1995). We obtained an unbiased, although potentially diluted, log‐rank analysis of head and neck cancer mortality indirectly by subtracting the log‐rank statistic for non‐head and neck cancer mortality from the log‐rank statistic for mortality from all causes (i.e. the two observed values are subtracted from each other, the two expected values are subtracted from each other, and the two variances are subtracted from each other).

We used Chi² tests to study heterogeneity between trials and between trial groups (NSCLC‐CG 1995). We used the I² statistic to estimate the proportion of variability of the results related to heterogeneity rather than to sampling error (Higgins 2002). To study the interaction between treatment and covariates, we carried out an analysis stratified on trials for each covariate value, and compared the HR for the different values of the covariate with a heterogeneity test. We computed stratified survival curves for control and experimental groups with annual death rates and HR, and then used these to calculate absolute benefit at two years and five years (EBCT‐CG 1992). All P values are two‐sided.

Results

Description of studies

Of 26 potentially eligible randomised trials we excluded nine: three were postoperative trials (Ang 2001; Awwad 1992; Awwad 2002), one had biased randomisation (Tandon 1999), two used unconventional radiotherapy in the reference group (Sanchiz 1990; Van den Bogaert 1986) and three used hypofractionated radiotherapy in the experimental group (Overgaard 1989; Weissberg 1982; Wiernik 1991). The 'ongoing' trials list (see the table Characteristics of ongoing studies) includes both trials that ended accrual of patients after the period of inclusion for trials defined in the protocol (1998) and trials still ongoing.

Data from one eligible trial (212 patients) were lost (Datta 1989). Fifteen trials fulfilled all the inclusion criteria and data were available for 6515 patients (BCCA 9113 1997; CAIR 2000; CHART 1997; DAHANCA 2003; EORTC 22791 1992; EORTC 22851 1997; GORTEC 9402 2006; KBN PO 79 2002; Oro 9301 2003; PMHToronto 2007; RIO 1991; RTOG 7913 1987; RTOG 9003HF 2000; RTOG 9003B 2000; RTOG 9003S 2000; TROG 9101 2001; Vienna 2000). The length of follow up varied from four to 10 years, with a median of six years. One trial, RTOG 9003, was a four‐arm trial, with a reference arm and three experimental arms. The reference arm of this trial was counted three times so that the three altered fraction modalities of the trial could be analysed separately (RTOG 9003B 2000; RTOG 9003HF 2000; RTOG 9003S 2000). Overall, 17 comparisons were made for 7073 patients.

Trial and patient characteristics are presented in Figure 1 and Figure 2.

1.

Description of trials comparing altered fractionated with conventional radiotherapy: patients characteristics

BCCA = British Columbia Cancer Agency; CAIR = Continuous Accelerated Irradiation; CHART = Continuous Hyperfractionated Accelerated Radiation Therapy; DAHANCA = Danish Head and Neck Cancer Study Group; EORTC = European Organisation for Research and Treatment of Cancer; GORTEC = Groupe d'Oncologie Radiothérapie Tête et Cou; KBN = Komiet Badan Naukowych (Committee for Scientific Research); PMH‐Toronto = Princess Margaret Hospital, Toronto; RTOG = Radiation Therapy Oncology Group; TROG = Trans‐Tansman Radiation Oncology Group

Figure from Bourhis J, Overgaard J, Audry H, Ang KK, Saunders M, Bernier J et al on behalf of MARCH collaborative group. Hyperfractionated or accelerated radiotherapy in head and neck cancer: a meta‐analysis. Lancet 2006;368:843‐54 reproduced with permission from Elsevier Ltd.

2.

Description of patients included in trials comparing altered fractionated radiotherapy with conventional radiotherapy by arm (n = 7073).

Figure from Bourhis J, Overgaard J, Audry H, Ang KK, Saunders M, Bernier J et al on behalf of MARCH collaborative group. Hyperfractionated or accelerated radiotherapy in head and neck cancer: a meta‐analysis. Lancet 2006;368:843‐54 reproduced with permission from Elsevier Ltd.

Risk of bias in included studies

All the included trials fulfilled the strict methodological criteria of this individual patient data meta‐analysis. One study was excluded because of biased randomisation (Tandon 1999). For each trial, the balance of covariates and the cumulative number of patients accrued over time were checked to identify possible problems with randomisation. Moreover, we compared the length of follow up by arm. Following these investigations, we identified no problems among the eligible trials. All trials scored 'A' for randomisation concealment. Some trials did not perform an intention‐to‐treat analysis but we were able to gather data for 154 of the 163 randomised patients who had been excluded from the published analyses. Lastly the length of follow up was good.

We also assessed study quality using the Cochrane 'Risk of bias' approach. The results of our assessments are shown in the 'Risk of bias' tables (Characteristics of included studies). A 'Risk of bias' graph (review authors' judgements about each risk of bias item presented as percentages across all included studies) and summary (review authors' judgements about each risk of bias item for each included study) are shown in Figure 3 and Figure 4.

3.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies. None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding.

4.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study. None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding.

Effects of interventions

There was a good balance between the trial arms for site, stage, sex, histology, age and performance status. The main tumour sites were oropharynx (3079 patients, 44%) and larynx (2377, 34%). In total, 1812 (26%) patients had stage I and II and 5221 (74%) had stage III‐IV tumours (UICC 2002).

There was a significant benefit for overall survival with altered fractionation radiotherapy compared with conventional radiotherapy (Analysis 1.1; Figure 5). This benefit corresponded to an 8% (95% CI 3 to 14) reduction in the risk of dying and an absolute benefit of 3.3% (0.9 to 5.7) and 3.4% (1.0 to 5.8) at two and five years, respectively.

1.1. Analysis.

Comparison 1 Altered fractionated radiotherapy versus conventional radiotherapy, Outcome 1 Hazard ratio of death.

5.

Survival curves by treatment arm for all trials and for the three groups of trials according to the type of altered fractionated radiotherapy. The slopes of the broken lines from year 6 to year >= 7 are based on the overall death rates in the seventh and subsequent years.

RT = radiotherapy

Figure from Bourhis J, Overgaard J, Audry H, Ang KK, Saunders M, Bernier J et al on behalf of MARCH collaborative group. Hyperfractionated or accelerated radiotherapy in head and neck cancer: a meta‐analysis. Lancet 2006;368:843‐54 reproduced with permission from Elsevier Ltd.

Heterogeneity was significant between trials (P = 0.001; I² = 58%). Altered fractionation radiotherapy had no effect on death not related to cancer (HR 1.06, 95% CI 0.93 to 1.22) and the overall benefit was due to the effect on death related to cancer (HR 0.88, 95% CI 0.83 to 0.94) (Analysis 1.2; Figure 6). The magnitude of the survival benefit was significantly higher in the hyperfractionation group than in the two other groups (test for interaction, P = 0.02; Analysis 1.1; Figure 5). This comparison should be interpreted with caution because the populations included in the three groups were dissimilar (Figure 7), for example more patients with early stage or larynx tumour were included in the group with accelerated fractionation and the same total dose.

1.2. Analysis.

Comparison 1 Altered fractionated radiotherapy versus conventional radiotherapy, Outcome 2 Hazard ratio of head and neck cancer death.

6.

Non‐cancer death and cancer death survival curves for all trials and for the three groups of trials according to the altered fractionated radiotherapy. The slopes of the broken lines from year 6 to year >= 7 are based on the overall death rates in the seventh and subsequent years.

RT = radiotherapy

Figure from Bourhis J, Overgaard J, Audry H, Ang KK, Saunders M, Bernier J et al on behalf of MARCH collaborative group. Hyperfractionated or accelerated radiotherapy in head and neck cancer: a meta‐analysis. Lancet 2006;368:843‐54 reproduced with permission from Elsevier Ltd.

7.

Description of patients included in trials comparing conventional radiotherapy with altered fractionated radiotherapy by group of altered fractionated radiotherapy (n = 7073).

Figure from Bourhis J, Overgaard J, Audry H, Ang KK, Saunders M, Bernier J et al on behalf of MARCH collaborative group. Hyperfractionated or accelerated radiotherapy in head and neck cancer: a meta‐analysis. Lancet 2006;368:843‐54 reproduced with permission from Elsevier Ltd.

Data about local and regional failures were available for only 14 of 17 trials (6410 patients; Figure 8). Local recurrence was the main cause of first failure (2527 patients (39%), isolated in 1544 patients (24%) and associated with only a regional neck lymph node failure in 909 (14%)), whereas regional failure was reported in 1407 (22%) patients (isolated in 419 (7%)). Finally, distant metastases were reported in 533 (8%) patients (isolated in 360 (5%)).

8.

Distribution of type of first failure by arm.

Figure from Bourhis J, Overgaard J, Audry H, Ang KK, Saunders M, Bernier J et al on behalf of MARCH collaborative group. Hyperfractionated or accelerated radiotherapy in head and neck cancer: a meta‐analysis. Lancet 2006;368:843‐54 reproduced with permission from Elsevier Ltd.

There was a significant benefit on locoregional control for altered fractionation compared with conventional radiotherapy (P < 0.0001; Analysis 1.3 and Figure 9). This benefit was seen in all three groups, but was slightly more pronounced in the two groups that did not decrease the total dose, compared with the reference arm (Analysis 1.3; Figure 9; Figure 10).

1.3. Analysis.

Comparison 1 Altered fractionated radiotherapy versus conventional radiotherapy, Outcome 3 Hazard ratio of locoregional control.

9.

Locoregional failure by treatment arm according to the type of radiotherapy. The slopes of the broken lines from year 6 to year >= 7 are based on the overall death rates in the seventh and subsequent years.

RT = radiotherapy

Figure from Bourhis J, Overgaard J, Audry H, Ang KK, Saunders M, Bernier J et al on behalf of MARCH collaborative group. Hyperfractionated or accelerated radiotherapy in head and neck cancer: a meta‐analysis. Lancet 2006;368:843‐54 reproduced with permission from Elsevier Ltd.

10.

Hazard ratio (95% CI) of altered fractionated radiotherapy versus conventional radiotherapy on overall population and type of radiotherapy for locoregional, local, regional, and metastatic control (n = 7073).

Figure from Bourhis J, Overgaard J, Audry H, Ang KK, Saunders M, Bernier J et al on behalf of MARCH collaborative group. Hyperfractionated or accelerated radiotherapy in head and neck cancer: a meta‐analysis. Lancet 2006;368:843‐54 reproduced with permission from Elsevier Ltd.

Altered fractionated radiotherapy was especially effective in the reduction of local failure in all three groups (Analysis 1.4), with a 23% reduction in the risk and an absolute benefit of 8.5% (5.7 to 11.3) at five years. The benefit of this treatment on regional control was also significant, with a 13% reduction in the risk and an absolute benefit of 1.9% (‐0.7 to 4.5) at five years, although much less pronounced than for local control (Figure 10). No effect of altered fractionation radiotherapy could be detected on distant metastases. The fact that the hyperfractionation group and the group with accelerated fractionation without total dose reduction shared the same benefit for local control (Analysis 1.4), but had a different effect on survival (Analysis 1.1; Figure 5), could be attributable to an excess of non‐cancer related deaths in the group with accelerated fractionation without total dose reduction (Figure 6). At five years, for example, 2.4% more patients in the group with accelerated fractionation without total dose reduction had non‐cancer related deaths.

1.4. Analysis.

Comparison 1 Altered fractionated radiotherapy versus conventional radiotherapy, Outcome 4 Hazard ratio of local control.

We performed several sensitivity analyses on overall survival, cancer mortality and locoregional control endpoints. These analyses excluded stage I and II tumours, the CAIR trial (a trial with outlying results), the Oro 9301 2003 and RTOG 9003HF 2000 trials (that did not fit perfectly in the group with accelerated fractionation without total dose reduction), and counted only once the control group of the RTOG 9003 trial (data not shown). These analyses led to very similar overall results and often to a decrease in the heterogeneity and its disappearance for all analyses excluding the CAIR trial.

There was no significant interaction between sex, performance status, tumour stage, nodal stage, overall stage, tumour site and the treatment effect on overall survival, but an interaction with age was recorded (Figure 11). Indeed, a test for trend revealed a significant interaction between age and treatment effect for overall survival (P = 0.007), and death related to cancer (P = 0.008), local control (P = 0.002) and locoregional control (P = 0.002). A significant interaction was also noted between performance status and treatment effect, but only for tumour control (test for trend, P < 0.0001 for locoregional control, P = 0.0001 for local control, P = 0.004 for regional control). The effect of altered fractionation radiotherapy on tumour control was higher in patients with good performance status. The effect of altered fractionation radiotherapy on tumour control did not differ significantly according to tumour stage and tumour site. Treatment effect on locoregional failure was better for N0 and N1 than for N2 or N3 nodal stage (test for trend P = 0.02).

11.

Hazard ratio of death with altered fractionated radiotherapy versus conventional radiotherapy by age, sex, performance status, stage and site of tumour. Test for trend was significant for age (P = 0.007).

Figure from Bourhis J, Overgaard J, Audry H, Ang KK, Saunders M, Bernier J et al on behalf of MARCH collaborative group. Hyperfractionated or accelerated radiotherapy in head and neck cancer: a meta‐analysis. Lancet 2006;368:843‐54 reproduced with permission from Elsevier Ltd.

Discussion

Individual patient data meta‐analysis is a long process, as the authors need to obtain all the updated data from the trialists before starting the analyses. MARCH (Meta‐Analysis of Radiotherapy in Carcinomas of Head and neck) started in 2000 and was first published in 2006 (Bourhis 2006). This is the reason why recent trials were not included in this meta‐analysis. MARCH 2, which will include more recent studies, is currently underway (see Implications for research).

Meta‐analysis of individual patient data shows that different types of altered fractionation radiotherapy could be more effective in treating head and neck squamous cell carcinomas, compared with conventional radiotherapy, with a small but significant benefit in survival and a more pronounced benefit in locoregional and local control. Our findings provide strong evidence that altered fractionation radiotherapy can improve survival in this disease. The survival benefit was mainly seen in the group with increased total dose (i.e. hyperfractionated radiotherapy) and corresponded to an absolute benefit of 8% at five years in this group. This benefit is of the same size as the effect due to the use of chemotherapy concomitantly with radiotherapy in this type of cancer (i.e. 8% at five years with the method used in a meta‐analysis of chemotherapy (Pignon 2000a; Stewart 1993) and 6.5% with the same method used here).

Analyses of the different types of altered fractionation radiotherapy suggest that the hyperfractionation group showed the greatest benefit (P = 0.02). However, this difference was noted only for survival, whereas for locoregional control a non‐significant trend only was recorded in favour of hyperfractionation and of accelerated fractionation without total dose reduction. These findings suggest that substantial acceleration could only partly compensate for decreasing the total dose (Analysis 1.3; Analysis 1.4; Figure 9). Increasing the total dose in hyperfractionated radiotherapy could be an attractive option, since this is the only group in which a benefit was seen both on survival and local control. However, the benefit in locoregional control was much the same in the group of trials with moderate acceleration and in which the total dose was kept the same as in the reference arm. It is necessary to define which characteristics of the patients and tumours could be used in order to select the optimum altered fractionation radiotherapy for individual patients.

The modest 3.4% survival benefit of altered fractionation radiotherapy at five years could be offset by an increased risk in late toxicities. Among the four trials reporting significant differences in late toxicity, two (BCCA 9113 1997; EORTC 22851 1997) showed an increased risk with accelerated fractionation without total dose reduction and two (CHART 1997; TROG 9101 2001) showed a decreased risk with accelerated fractionation with total dose reduction.

The effect of altered fractionation was significantly more pronounced on the primary tumour than on nodal disease. The interpretation of this observation is not easy, since we studied only the first site of failure and simultaneous failures in the primary site and in the nodes were frequent. Altered fractionation radiotherapy could be appropriate for patients with N0 and N1 disease, whereas combinations of chemotherapy and radiotherapy could be more appropriate for patients with more advanced nodal disease. Altered fractionation had no effect on distant metastases (Figure 10). However, this result should be viewed with caution as the low observed rate of distant metastases could be related to poor recording, thus resulting in low power for this analysis. In the patients randomly assigned to the conventional radiotherapy group, the overall survival of the larynx subgroup was significantly better than that for the other sites (data not shown). However, the effect of altered fractionation radiotherapy did not differ for the larynx compared with the other sites. A stratification of data on the larynx site did not change the results (data not shown).

The strong suggestion of a decreasing effect of altered fractionation radiotherapy with increasing age and with poor performance status might be partly explained by an excess of non‐cancer related deaths in patients aged 71 years and over (Pignon 2007), but also by lower compliance and tolerance in these patients and in patients with poor general health status (Khalil 2003). However, tolerance was difficult to assess from our database, since recording and scoring acute and late radiation effects could vary between trials. The decreasing effect of more intense treatment in older patients has also been reported in patients with head and neck squamous cell carcinomas treated with concomitant chemotherapy and radiotherapy (Bourhis 2004; Pignon 2000a).

Authors' conclusions

Implications for practice.

We have shown that altered fractionation radiotherapy confers greater benefit than conventional radiotherapy in tumour control and survival. The effect was greater for the primary tumour than for nodal disease. The effect was also more pronounced in younger patients and in those with good performance status. Hyperfractionation seemed to yield a more consistent advantage for survival than accelerated radiotherapy. However, there was more diversity in accelerated fractionation regimens than in hyperfractionated regimens, and some of these regimens might be associated with higher non‐cancer related death, off‐setting their benefit in improving tumour control.

Implications for research.

An update of MARCH is scheduled (MARCH 2). This study should increase the power of this analyses and allow for other comparisons: current trials investigate whether the benefit of hyperfractionated radiotherapy versus standard radiotherapy persists when combined with concomitant chemotherapy or when applied in a postoperative setting. MARCH 2 may allow us to provide a global estimator of the treatment effect. Strategies with intensity‐modulated radiotherapy or targeted therapy will also have to be assessed.

What's new

Date	Event	Description
3 April 2015	Amended	Amendment to Declarations of interest section.

History

Protocol first published: Issue 2, 2000
 Review first published: Issue 12, 2010

Date	Event	Description
26 September 2008	Amended	Converted to new review format.

Notes

The financial sponsors of this study had no role in the study design, data collection, data analysis, data interpretation or the writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Appendices

Appendix 1. Search strategies

CENTRAL	PubMed	EMBASE (Ovid)	CINAHL (EBSCO)
#1 MeSH descriptor Head and Neck Neoplasms, this term only #2 MeSH descriptor Otorhinolaryngologic Neoplasms explode all trees #3 MeSH descriptor Neoplasms explode all trees #4 (cancer* or carcinom* or tumor* or tumour* or neoplas*):ti #5 (#3 OR #4) #6 MeSH descriptor Larynx explode all trees #7 MeSH descriptor Pharynx explode all trees #8 MeSH descriptor Mouth explode all trees #9 (HNSCC or SCCHN or head NEAR neck or oral NEXT cavity or oropharyn* or hypopharyn* or laryn*):ti #10 (#6 OR #7 OR #8 OR #9) #11 (#5 AND #10) #12 (#1 OR #2 OR #11) #13 MeSH descriptor Radiotherapy explode all trees #14 (Hyperfractionat* or Fraction* or Radiotherap* or Radiat* or Irradiat* or CFRT or CTRT or ACRT or AHF or CF or ARTSCAN or CHARTWELL or DAHANCA or EORTC or GORTEC or POPART or PARTIR or RTOG):ti #15 (#13 OR #14) #16 (#12 AND #15) #17 (conventional* or cfrt or cf or ((standard or classic*) and (fraction* or hyperfraction* or superfraction* or radio*))) #18 (#16 AND #17)	#1 "Head and Neck Neoplasms"[Mesh:NoExp] OR "Otorhinolaryngologic Neoplasms"[Mesh] #2 "Neoplasms"[Mesh] OR cancer* [ti] OR carcinom* [ti] OR tumor* [ti] OR tumour* [ti] OR neoplas* [ti] #3 “Larynx” [Mesh] OR “Pharnyx” [Mesh] OR HNSCC [ti] OR SCCHN [ti] OR “head and neck” [ti] OR “head neck” [ti] OR “head‐neck” [ti] OR “head‐and‐neck” [ti] OR "oral cavity" [ti] OR oropharyn* [ti] OR hypopharyn* [ti] OR laryn* [ti] #4 #2 AND #3 #5 #1 OR #4 #6 "Radiotherapy"[Mesh] OR "Radiotherapy"[subheading] OR Hyperfractionat* [ti] OR Fraction* [ti] OR Radiotherap* [ti] OR Radiat* [ti] OR Irradiat* [ti] OR CFRT [ti] OR CTRT [ti] OR ACRT [ti] OR AHF [ti] OR CF [ti] OR ARTSCAN[ti] OR CHARTWELL [ti] OR DAHANCA [ti] OR EORTC [ti] OR GORTEC [ti] OR POPART [ti] OR PARTIR [ti] OR RTOG [ti] #7 #5 AND #6 #8 conventional* [tiab] OR cfrt [tiab] OR cf [tiab] OR ((standard [tiab] OR classic* [tiab]) AND (fraction* [tiab] OR hyperfraction* [tiab] OR superfraction* [tiab] OR radio* [tiab])) #9 #7 AND #8	1 exp *"head and neck tumor"/ 2 exp larynx tumor/ 3 exp neoplasm/ 4 (cancer* or carcinom* or tumor* or tumour* or neoplas*).ti. 5 3 or 4 6 exp mouth cavity/ 7 exp pharynx/ 8 exp larynx/ 9 (HNSCC or SCCHN or (head and neck) or "oral cavity" or oropharyn* or hypopharyn* or laryn*).ti. 10 6 or 7 or 8 or 9 11 5 and 10 12 1 or 2 or 11 13 exp radiotherapy/ 14 (Hyperfractionat* or Fraction* or Radiotherap* or Radiat* or Irradiat* or CFRT or CTRT or ACRT or AHF or CF or ARTSCAN or CHARTWELL or DAHANCA or EORTC or GORTEC or POPART or PARTIR or RTOG).ti. 15 13 or 14 16 12 and 15 17 (conventional* or cfrt or cf or ((standard or classic*) and (fraction* or hyperfraction* or superfraction* or radio*))).tw. 18 16 AND 17	S1 (MH "Otorhinolaryngologic Neoplasms+") S2 (MM "Head and Neck Neoplasms") S3 TI cancer* or carcinom* or tumor* or tumour* or neoplas* S4 TI HNSCC or SCCHN or (head and neck) or "oral cavity" or oropharyn* or hypopharyn* or laryn* S5 S3 and S4 S6 (MH "Radiotherapy+") S7 TI Hyperfractionat* OR Fraction* OR Radiotherap* OR Radiat* OR Irradiat* OR CFRT OR CTRT OR ACRT OR AHF S8 S6 or S7 S9 S1 or S2 or S5 S10 S8 and S9 S11 TX conventional* or cfrt or standard or classic* S12 S10 and S11
Web of Science	BIOSIS Previews (Web of Knowledge)	CAB Abstracts (Ovid)	ISRCTN
#1 TI=((cancer* OR carcinom* OR neoplas* OR tumor* OR tumour*) AND (Radiotherap* OR Radiat* OR Irradiat* OR Hyperfractionat* OR Fraction* OR superfraction* OR ACRT OR AHF OR CF OR ARTSCAN OR CHARTWELL OR DAHANCA OR EORTC OR GORTEC OR POPART OR PARTIR OR RTOG)) #2 TI=(HNSCC or SCCHN or (head and neck) or "oral cavity" or oropharyn* or hypopharyn* or laryn*) #3 TS=(conventional* OR cfrt OR cf OR ((standard* OR classic*) AND (fraction* OR hyperfraction* OR superfraction* OR radio*))) #4 #3 AND #2 AND #1	#1 TI=((cancer* OR carcinom* OR neoplas* OR tumor* OR tumour*) AND (Radiotherap* OR Radiat* OR Irradiat* OR Hyperfractionat* OR Fraction* OR superfraction* OR ACRT OR AHF OR CF OR ARTSCAN OR CHARTWELL OR DAHANCA OR EORTC OR GORTEC OR POPART OR PARTIR OR RTOG)) #2 TI=(HNSCC or SCCHN or (head and neck) or "oral cavity" or oropharyn* or hypopharyn* or laryn*) #3 TS=(conventional* OR cfrt OR cf OR ((standard* OR classic*) AND (fraction* OR hyperfraction* OR superfraction* OR radio*))) #4 #3 AND #2 AND #1	1 HNSCC OR SCCHN 2 (cancer* or carcinom* or tumor* or tumour* or neoplas*).ti. 3 ((head and neck) or "oral cavity" or oropharyn* or hypopharyn* or laryn*).ti. 4 2 AND 3 5 1 OR 4 6 (Hyperfractionat* or Fraction* or Radiotherap* or Radiat* or Irradiat* or CFRT or CTRT or ACRT or AHF or CF or ARTSCAN or CHARTWELL or DAHANCA or EORTC or GORTEC or POPART or PARTIR or RTOG).ti. 7 (conventional* or cfrt or cf or ((standard or classic*) and (fraction* or hyperfraction* or superfraction* or radio*))).tw. 8 5 AND 6 AND 7	(conventional OR classic OR standard OR CF OR CFRT) AND radiotherapy (conventional OR classic OR standard OR CF OR CFRT) AND radiation AND therapy

Data and analyses

Comparison 1. Altered fractionated radiotherapy versus conventional radiotherapy.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Hazard ratio of death	17	7073	Peto Odds Ratio (95% CI)	0.92 [0.86, 0.97]
1.1 Hyperfractionation	4	1350	Peto Odds Ratio (95% CI)	0.78 [0.69, 0.89]
1.2 Accelerated fractionation without total dose reduction	8	3818	Peto Odds Ratio (95% CI)	0.97 [0.89, 1.05]
1.3 Accelerated fractionation with total dose reduction	5	1905	Peto Odds Ratio (95% CI)	0.94 [0.84, 1.05]
2 Hazard ratio of head and neck cancer death	17	7073	Peto Odds Ratio (95% CI)	0.88 [0.83, 0.94]
2.1 Hyperfractionation	4	1350	Peto Odds Ratio (95% CI)	0.78 [0.68, 0.90]
2.2 Accelerated fractionation without total dose reduction	8	3818	Peto Odds Ratio (95% CI)	0.91 [0.83, 1.00]
2.3 Accelerated fractionation with total dose reduction	5	1905	Peto Odds Ratio (95% CI)	0.93 [0.83, 1.05]
3 Hazard ratio of locoregional control	17	7073	Peto Odds Ratio (95% CI)	0.82 [0.77, 0.88]
3.1 Hyperfractionation	4	1350	Peto Odds Ratio (95% CI)	0.76 [0.66, 0.89]
3.2 Accelerated fractionation without total dose reduction	8	3818	Peto Odds Ratio (95% CI)	0.79 [0.72, 0.87]
3.3 Accelerated fractionation with total dose reduction	5	1905	Peto Odds Ratio (95% CI)	0.90 [0.80, 1.02]
4 Hazard ratio of local control	14	6410	Peto Odds Ratio (95% CI)	0.77 [0.71, 0.83]
4.1 Hyperfractionation	3	1247	Peto Odds Ratio (95% CI)	0.75 [0.63, 0.89]
4.2 Accelerated fractionation without total dose reduction	8	3818	Peto Odds Ratio (95% CI)	0.74 [0.67, 0.83]
4.3 Accelerated fractionation with total dose reduction	3	1345	Peto Odds Ratio (95% CI)	0.83 [0.71, 0.96]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

BCCA 9113 1997.

Methods	Inclusion period: 1991 to 1995 Median follow up: 7.8 years
Participants	82 patients
Interventions	Altered fractionated arm: 66 Gy, 2 daily fractions, 3.5 weeks Conventional arm: 66 Gy, 1 daily fraction, 6.5 weeks
Outcomes	Overall survival, locoregional control, local control, regional control, distant control, cause‐specific mortality
Notes	BCCA = British Columbia Cancer Agency Patients' written consent: yes / ethical board: unspecified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

CAIR 2000.

Methods	Inclusion period: 1994 to 1996 Median follow up: 5.7 years
Participants	100 patients
Interventions	Altered fractionated arm: 64 to 74 Gy, 1 daily fraction, 4.5 to 5 weeks Conventional arm: 64 to 74 Gy, 1 daily fraction, 6.5 to 8 weeks
Outcomes	Overall survival, locoregional control, local control, regional control, distant control, cause‐specific mortality
Notes	Because of the high incidence of toxicity in the accelerated arm, the schedules were modified for the next patients. The dose per fraction was reduced from 2.0 to 1.8 Gy and a second fraction was given twice a week in both arms. In the non‐accelerated arm, there was a 3‐day break each weekend CAIR = Continuous Accelerated Irradiation Patients' written consent/ethical board: yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

CHART 1997.

Methods	Inclusion period: 1990 to 1995 Median follow up: 7.0 years
Participants	918 patients
Interventions	Altered fractionated arm: 54 Gy, 3 daily fractions, 1.7 weeks Conventional arm: 66 Gy, 1 daily fraction, 6.5 weeks
Outcomes	Overall survival, locoregional control, local control, regional control, distant control, cause‐specific mortality
Notes	CHART = Continuous Hyperfractionated Accelerated Radiation Therapy Patients' written consent/ethical board: yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

DAHANCA 2003.

Methods	Inclusion period: 1991 to 1999 Median follow up: 6.8 years
Participants	1485 patients
Interventions	Altered fractionated arm: 66 to 68 Gy, 1 daily fraction, 6 weeks Conventional arm: 66 to 68 Gy, 1 daily fraction, 7 weeks
Outcomes	Overall survival, locoregional control, local control, regional control, distant control, cause‐specific mortality
Notes	5.5% of patients included in 1999. Radiosensitiser nimorazole in both arms for DAHANCA 7 (791 patients) DAHANCA = Danish Head and Neck Cancer Study Group Patients' written consent/ethical board: yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

EORTC 22791 1992.

Methods	Inclusion period: 1980 to 1987 Median follow up: 10.3 years
Participants	356 patients
Interventions	Altered fractionated arm: 80.5 Gy, 2 daily fractions, 7 weeks Conventional arm: 70 Gy, 1 daily fraction, 7 weeks
Outcomes	Overall survival, locoregional control, local control, regional control, distant control, cause‐specific mortality
Notes	EORTC = European Organisation for Research and Treatment of Cancer Patients' written consent/ethical board: unspecified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

EORTC 22851 1997.

Methods	Inclusion period: 1985 to 1995 Median follow up: 4.8 years
Participants	512 patients
Interventions	Altered fractionated arm: 72 Gy, 3 daily fractions, 5 weeks sc‡ Conventional arm: 70 Gy, 1 daily fraction, 7 weeks
Outcomes	Overall survival, locoregional control, local control, regional control, distant control, cause‐specific mortality
Notes	‡First course: 28.8 Gy, in 18 fractions for 8 days, 12 to 14 days split, 43.2 Gy in 27 fractions for 17 days EORTC = European Organisation for Research and Treatment of Cancer Patients' written consent/ethical board: unspecified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

GORTEC 9402 2006.

Methods	Inclusion period: 1994 to 1998 Median follow up: 4.8 years
Participants	268 patients
Interventions	Altered fractionated arm: 62 to 64 Gy, 2 daily fractions, 3 weeks Conventional arm: 70 Gy, 1 daily fraction, 7 weeks
Outcomes	Overall survival, locoregional control, local control, regional control, distant control, cause‐specific mortality
Notes	GORTEC = Groupe d'Oncologie Radiothérapie Tête et Cou Patients' written consent: yes/ethical board: unspecified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

KBN PO 79 2002.

Methods	Inclusion period: 1995 to 1998 Median follow up: 4.1 years
Participants	395 patients
Interventions	Altered fractionated arm: 66 Gy, 1 daily fraction, 5.5 weeks b Conventional arm: 66 Gy, 1 daily fraction, 6.5 weeks
Outcomes	Overall survival, locoregional control, local control, regional control, distant control, cause‐specific mortality
Notes	KBN = Komiet Badan Naukowych (Committee for Scientific Research) Patients' written consent/ethical board: yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

Oro 9301 2003.

Methods	Inclusion period: 1993 to 1998 Median follow up: 6.6 years
Participants	128 patients
Interventions	Altered fractionated arm: 64 to 67 Gy, 2 daily fractions, 6.5 weeks sc* Conventional arm: 66 to 70 Gy, 1 daily fraction, 7 weeks
Outcomes	Overall survival, locoregional control, local control, regional control, distant control, cause‐specific mortality
Notes	Third arm with radio‐chemotherapy *First course: 38.4 Gy in 24 fractions for 2.5 weeks, 2 weeks split, 25.6 to 28.8 Gy in 16 to 18 fractions for 2.5 to 3 weeks Patients' written consent: yes/ethical board: unspecified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

PMHToronto 2007.

Methods	Inclusion period: 1988 to 1995 Median follow up: 7.4 years
Participants	336 patients
Interventions	Altered fractionated arm: 58 Gy, 2 daily fractions of 1.45 Gy, 4 weeks Conventional arm: 51 Gy, 1 daily fraction 2.5 Gy, 4 weeks**
Outcomes	Overall survival, locoregional control, local control, regional control, distant control, cause‐specific mortality
Notes	** 51 Gy translates to 50 Gy prescribed at the 90% isodose It was considered by the investigators that 20 fraction of 2.5 Gy over 28 days was equivalent to 66 to 70 Gy using 2 Gy daily fractions over 6.5 to 7 weeks PMH‐Toronto = Princess Margaret Hospital, Toronto Patients' written consent/ethical board: yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

RIO 1991.

Methods	Inclusion period: 1986 to 1989 Median follow up: 6.7 years
Participants	103 patients
Interventions	Altered fractionated arm: 70.4 Gy, 2 daily fractions, 6.5 weeks Conventional arm: 66 Gy, 1 daily fraction, 6.5 weeks
Outcomes	Overall survival, locoregional control, distant control, cause‐specific mortality
Notes	Patients' written consent/ethical board: yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

RTOG 7913 1987.

Methods	Inclusion period: 1979 to 1983 Median follow up: 9.2 years
Participants	210 patients
Interventions	Altered fractionated arm: 60 Gy, 2 daily fractions, 6 weeks Conventional arm: 70 Gy, 1 daily fraction, 7 weeks
Outcomes	Overall survival, locoregional control, distant control, cause‐specific mortality
Notes	RTOG = Radiation Therapy Oncology Group Patients' written consent: yes/ethical board: unspecified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

RTOG 9003B 2000.

Methods	Inclusion period: 1991 to 1997 Median follow up: 6.0 years
Participants	556 patients
Interventions	Altered fractionated arm: 72 Gy, 2 daily fractions, 6 weeks b Conventional arm: 70 Gy, 1 daily fraction, 7 weeks
Outcomes	Overall survival, locoregional control, local control, regional control, distant control, cause‐specific mortality
Notes	Four‐arm trials, each experimental arm was compared with the control arm RTOG = Radiation Therapy Oncology Group 'RTOG 9003 B' for boost Patients' written consent/ethical board: yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

RTOG 9003HF 2000.

Methods	Inclusion period: 1991 to 1997 Median follow up: 6.0 years
Participants	555 patients
Interventions	Altered fractionated arm: 81.6 Gy, 2 daily fractions, 7 weeks Conventional arm: 70 Gy, 1 daily fraction, 7 weeks
Outcomes	Overall survival, locoregional control, local control, regional control, distant control, cause‐specific mortality
Notes	Four‐arm trials, each experimental arm was compared with the control arm RTOG = Radiation Therapy Oncology Group 'RTOG 9003 HF' for hyperfractionated Patients' written consent/ethical board: yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

RTOG 9003S 2000.

Methods	Inclusion period: 1991 to 1997 Median follow up: 6.0 years
Participants	560 patients
Interventions	Altered fractionated arm: 67.6 Gy, 2 daily fractions, 6 weeks sc Conventional arm: 70 Gy, 1 daily fraction, 7 weeks
Outcomes	Overall survival, locoregional control, local control, regional control, distant control, cause‐specific mortality
Notes	Four‐arm trials, each experimental arm was compared with the control arm RTOG = Radiation Therapy Oncology Group 'RTOG 9003 S' for split course Patients' written consent/ethical board: yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

TROG 9101 2001.

Methods	Inclusion period: 1991 to 1998 Median follow up: 3.9 years
Participants	350 patients
Interventions	Altered fractionated arm: 59.4 Gy, 2 daily fractions, 3.5 weeks Conventional arm: 70 Gy, 1 daily fraction, 7 weeks
Outcomes	Overall survival, locoregional control, distant control, cause‐specific mortality
Notes	TROG = Trans‐Tansman Radiation Oncology Group Patients' written consent/ethical board: yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

Vienna 2000.

Methods	Inclusion period: 1990 to 1997 Median follow up: 5.6 years
Participants	159 patients
Interventions	Altered fractionated arm: 55 Gy, 2 daily fractions, 2.5 weeks Conventional arm: 70 Gy, 1 daily fraction, 7 weeks
Outcomes	Overall survival, locoregional control, local control, regional control, distant control, cause‐specific mortality
Notes	Third arm with accelerated radiotherapy plus mitomycin C Patients' written consent/ethical board: yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Unclear risk	None of the studies of altered fractionation used a sham procedure for radiotherapy blinding, but overall survival is a robust endpoint that should not be sensitive to this absence of blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk
Selective reporting (reporting bias)	Low risk
Other bias	Low risk

b = boost; Gy = Gray; sc = split course

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Awwad 2002 radical RT	ALLOCATION: biased randomisation, incomplete data
Datta 1989	Data were lost
Overgaard 1989	ALLOCATION: randomised trial PARTICIPANTS: 626 patients with pharynx and larynx carcinoma INTERVENTIONS: hypofractionated radiotherapy in the experimental group
Sanchiz 1990	ALLOCATION: randomised trial PARTICIPANTS: 859 patients with T3‐T4, NO‐3, MO head and neck cancer INTERVENTIONS: unconventional radiotherapy in the reference group
Skladowski 2007	ALLOCATION: randomised trial INTERVENTIONS: no standard arm (CAIR versus concomitant boost)
Tandon 1999	ALLOCATION: biased randomisation
Van den Bogaert 1986	ALLOCATION: randomised trial PARTICIPANTS: 523 patients with advanced head and neck cancer INTERVENTIONS: unconventional radiotherapy in the reference group
Weissberg 1982	ALLOCATION: randomised trial PARTICIPANTS: 64 patients with non‐resectable head and neck cancer INTERVENTIONS: hypofractionated radiotherapy in the experimental group
Wiernik 1991	ALLOCATION: randomised trial PARTICIPANTS: 611 patients with locally advanced head and neck cancer INTERVENTIONS: hypofractionated radiotherapy in the experimental group

CAIR = Continuous Accelerated Irradiation

Characteristics of ongoing studies [ordered by study ID]

Ang 2001.

Trial name or title	Randomized trial addressing risk feature and times factors of surgery plus radiotherapy in advanced head and neck cancer
Methods	Randomised trial
Participants	151 patients
Interventions	Conventional postoperative RT versus accelerated postoperative RT
Outcomes	Survival, LRC
Starting date	Inclusion period: 1991 to 1997
Contact information	See reference
Notes	None

Ang 2010.

Trial name or title	RTOG H0129
Methods	Randomised trial
Participants	720 patients
Interventions	Conventional RT + cisplatin versus accelerated fractionation by concomitant boost + cisplatin
Outcomes	Survival, LRC, DFS
Starting date	Inclusion period: NA
Contact information	NA
Notes	None

ARTSCAN II.

Trial name or title	ARTSCAN II
Methods	Randomised trial
Participants	260 patients planned (resectable cancer of the oral cavity)
Interventions	Preoperative accelerated versus postoperative conventional radiotherapy in patients with resectable cancer of the oral cavity
Outcomes	Primary: local and regional tumour control Secondary: OS, DFS, quality of life
Starting date	2008
Contact information	Johan Wennerberg (Dept of ORL/Head & Neck Surgery University Hospital Lund, SE‐22185, Lund, Sweden)
Notes	Potentially confounded trial which studies both a fractionation modification and a difference in treatment strategy (preoperative versus postoperative RT)

Awwad 1992.

Trial name or title	Accelerated versus conventional fractionation in the postoperative irradiation of locally advanced head and neck cancer: influence of tumour proliferation
Methods	Randomised trial
Participants	56 patients
Interventions	Conventional RT versus accelerated hyperfractionated RT
Outcomes	Survival, DFS
Starting date	Inclusion period: 1987 to 1989
Contact information	See reference
Notes	None

Awwad 2002.

Trial name or title	Accelerated hyperfractionation (AHF) compared to conventional fractionation (CF) in the postoperative radiotherapy of locally advanced head and neck cancer: influence of proliferation
Methods	Randomised trial
Participants	70 patients
Interventions	Conventional RT versus accelerated hyperfractionation RT
Outcomes	Survival, LRC, DFS
Starting date	Inclusion period: 1995 to 1997
Contact information	See reference
Notes	None

Bartelink 2002.

Trial name or title	Concomitant cisplatin and radiotherapy in a conventional and modified fractionation schedule in locally advanced head and neck cancer: a randomised phase II EORTC trial
Methods	Randomised trial
Participants	53 patients
Interventions	Conventional RT + concomitant cisplatin versus multiple fractions per day RT + concomitant cisplatin
Outcomes	Survival, LRC
Starting date	Inclusion period: NA
Contact information	See reference
Notes	None

Bourhis 2010.

Trial name or title	GORTEC 99‐02
Methods	Randomised trial
Participants	840 patients
Interventions	Conventional RT + concomitant CT (5FU‐carboplatin) versus accelerated RT + concomitant CT (5FU‐carboplatin) versus very accelerated RT
Outcomes	Survival, DFS
Starting date	Inclusion period: 2000 to NA
Contact information	NA
Notes	Trial with 3 arms

Dische 2007.

Trial name or title	CHARTWEL
Methods	Randomised trial
Participants	460 patients
Interventions	Conventional RT versus CHARTWEL (continuous hyperfractionated accelerated RT)
Outcomes	Survival, DFS
Starting date	Inclusion period: 2001 to 2004
Contact information	See reference
Notes	None

Ezzat 2005.

Trial name or title	Randomized study of accelerated fractionation radiotherapy with and without mitomycin C in the treatment of locally advanced head and neck cancer
Methods	Randomised trial
Participants	40 patients
Interventions	Conventional fractionation RT versus accelerated fractionation RT
Outcomes	Survival, LRC
Starting date	Inclusion period: 1998 to 2001
Contact information	See reference
Notes	Trial with 3 arms but only 2 arms are eligible

Ghosh 2006.

Trial name or title	Randomized trial of conventional fractionated RT (CFRT) vs concomitant chemo radiotherapy (CTRT) and accelerated radiotherapy (ACRT) in patients with advanced, non nasopharyngeal, squamous cell cancers of the head and neck region
Methods	Randomised trial
Participants	150 patients
Interventions	Conventional RT + concomitant cisplatin versus hyperfractionated RT + concomitant cisplatin
Outcomes	Survival, DFS
Starting date	Inclusion period: 2000 to 2004
Contact information	See reference
Notes	Trial with 3 arms

Horiot 2007.

Trial name or title	EORTC 22962
Methods	Randomised trial
Participants	57 patients
Interventions	Conventional RT versus hyperfractionated RT
Outcomes	Survival, DFS
Starting date	Inclusion period: 1996 to 1999
Contact information	See reference
Notes	None

Johnson 1995.

Trial name or title	Standard once daily versus thrice‐daily concomitant boost accelerated superfractionated irradiation for advanced squamous cell carcinoma of the head and neck: preliminary results of a prospective randomized trial
Methods	Randomised trial
Participants	34 patients
Interventions	Conventional RT versus thrice‐daily concomitant boost
Outcomes	Survival, LRC
Starting date	Inclusion period: 1992 to NA
Contact information	See reference
Notes	None

Krstevska 2006.

Trial name or title	Altered and conventional fractionated radiotherapy in locoregional control and survival of patients with squamous cell carcinoma of the larynx, oropharynx, and hypopharynx
Methods	Randomised trial
Participants	152 patients
Interventions	Conventional RT versus hyperfractionation RT versus accelerated fractionation RT
Outcomes	Survival, LRC
Starting date	Inclusion period: 1999 to 2004
Contact information	See reference
Notes	Trial with 3 arms

Langendijk 2007a.

Trial name or title	POPART
Methods	Randomised trial
Participants	350 patients
Interventions	Conventional postoperative RT versus accelerated postoperative RT
Outcomes	Survival, LRC
Starting date	Inclusion period: Ongoing
Contact information	See reference
Notes	See reference

Langendijk 2007b.

Trial name or title	PARTIR
Methods	Randomised trial
Participants	360 patients
Interventions	Conventional postoperative RT versus accelerated postoperative RT
Outcomes	Survival, LRC
Starting date	Inclusion period: Ongoing
Contact information	See reference
Notes	See reference

Overgaard 2010.

Trial name or title	IEAE‐ACC
Methods	Randomised trial
Participants	908 patients
Interventions	Conventional RT versus 6 weekly fractions RT
Outcomes	Survival, LRC, DFS
Starting date	Inclusion period: 1999 to 2004
Contact information	See reference
Notes	None

Sanguineti 2005.

Trial name or title	Accelerated versus conventional fractionated postoperative radiotherapy for advanced head and neck cancer: results of a multicenter phase III study
Methods	Randomised trial
Participants	226 patients
Interventions	Conventional RT versus accelerated fractionated RT
Outcomes	Survival, LRC
Starting date	Inclusion period: 1994 to 2000
Contact information	See reference
Notes	None

Sastri 2008.

Trial name or title	Concomitant boost: an effective regimen in locoregionally advanced head and neck cancers ‐ a phase III randomised trial from a single institute in India
Methods	Randomised trial
Participants	285 patients
Interventions	Conventional RT versus concomitant boost RT
Outcomes	Survival, DFS
Starting date	Inclusion period: NA
Contact information	See reference
Notes	None

Slevin 2002.

Trial name or title	A randomised two arm trial of modestly accelerated radiotherapy with synchronous cisplatinum chemotherapy versus conventional radiotherapy with synchronous cisplatinum chemotherapy in the treatment of head and neck squamous cell carcinoma: a pilot study
Methods	Randomised trial
Participants	12 patients
Interventions	Accelerated or conventional radiotherapy with concomitant cisplatinum (100 mg/m2 every 3 weeks)
Outcomes	Tolerability, DFS, toxicity
Starting date	2002
Contact information	NJ Slevin (Clinical Oncology Christie Hospital NHS Trust, Wilmslow Road, Withington, M20 4BX, Manchester, United Kingdom)
Notes	None

Suwinski 2008.

Trial name or title	p‐CAIR
Methods	Randomised trial
Participants	279 patients
Interventions	Conventional postoperative RT versus 7 days a week postoperative continuous RT
Outcomes	Survival, LRC
Starting date	Inclusion period: 2001 to 2004
Contact information	See reference
Notes	None

Trotti 2006.

Trial name or title	RTOG 9512
Methods	Randomised trial
Participants	250 patients
Interventions	Conventional RT versus hyperfractionated RT
Outcomes	Survival, LRC, DFS
Starting date	Inclusion period: 1996 to 2001
Contact information	See reference
Notes	None

Zackrisson 2007.

Trial name or title	ARTSCAN
Methods	Randomised trial
Participants	750 patients
Interventions	Conventional RT versus accelerated fractionation RT with a concomitant boost
Outcomes	Survival, LRC
Starting date	Inclusion period: 1998 to 2001
Contact information	See reference
Notes	None

AHF = accelerated hyperfractionation
 DFS = disease‐free survival
 EORTC = European Organisation for Research and Treatment of Cancer
 LRC = locoregional control
 NA = not available
 OS = overall survival
 RT = radiotherapy
 RTOG = Radiation Therapy Oncology Group

This table presents the characteristics of studies that ended after the period of inclusion for trials defined in the protocol (1998) and studies still ongoing.

Differences between protocol and review

We have adopted and used the Cochrane 'Risk of bias' method for the assessment of study quality.

Contributions of authors

B Baujat, J Bourhis, J‐P Pignon, L Duchateau, R Sylvester and M Bolla, with the help of the members of the steering committee, contributed to the conception of the study. J‐P Pignon, N Syz, A Le Maître, L Duchateau and R Sylvester collected and checked the data with the help of the investigators who validated the re‐analysis of their trials. J‐P Pignon, B Baujat, A Le Maître and E Maillard did the statistical analysis. The manuscript was drafted by B Baujat, J Bourhis, J Overgaard, K Ang, A Le Maître, P Blanchard, E Maillard and J‐P Pignon and submitted for comments to the members of the secretariat and the steering committee. The investigators contributed to the interpretation of the results during the investigator meeting and revision of the manuscript. All authors have seen and approved the final version.

H Andrzej died prior to publication of this review. Substantial changes to the review were made beyond his contribution.

Sources of support

Internal sources

• Institut Gustave‐Roussy, France.

External sources

• Association pour la Recherche sur le Cancer n°5137, France.

• Programme Hospitalier de Recherche Clinique n°IDF98083, France.

• Ligue Nationale Contre le Cancer, France.

• Sanofi Aventis unrestricted grant, France.

• US National Cancer Institute 2U10CA11488‐36, USA.

Declarations of interest

Sanofi‐Aventis had a general agreement with Institut Gustave Roussy to support meta‐analyses on lung and head and neck cancers. Sanofi‐Aventis played no role in the study design, data collection, data analysis, data interpretation or writing of the publications. Sanofi‐Aventis is a pharmaceutical company with, to the best of the authors' knowledge, no interest in radiotherapy material.

We declare that we have no conflict of interest.

H Andrzej is deceased; no declaration of interest available.

Edited (no change to conclusions)