Radiotherapy and Oncology 100 (2011) 33–40

Contents lists available at ScienceDirect

Radiotherapy and Oncology

journal homepage: www.thegreenjournal.com

Meta-analysis of radiotherapy in HNSCC

Meta-analysis of chemotherapy in head and neck cancer (MACH-NC):

A comprehensive analysis by tumour site
Pierre Blanchard a,b,1, Bertrand Baujat c,1, Victoria Holostenco a, Abderrahmane Bourredjem a,
Charlotte Baey a, Jean Bourhis b, Jean-Pierre Pignon b,⇑, on behalf of the MACH-CH Collaborative group 2
a
Biostatistics and Epidemiology Department; and b Radiotherapy Department, Institut Gustave Roussy, Villejuif, France; c Head and Neck Surgery, Hôpital Foch, Suresnes, France

a r t i c l e i n f o a b s t r a c t

Article history: Introduction: The recently updated meta-analysis of chemotherapy in head and neck cancer (MACH-NC)
Received 14 March 2011 demonstrated the benefit of the addition of chemotherapy in terms of overall survival in head and neck
Received in revised form 17 May 2011 squamous cell carcinoma (HNSCC). The magnitude of the benefit according to tumour site is unknown as
Accepted 17 May 2011
well as their potential interactions with patient or trial characteristics.
Available online 16 June 2011
Methods: Eighty seven randomized trials performed between 1965 and 2000 were included in the pres-
ent analysis. Patients were divided into four categories according to tumour location: oral cavity, oro-
Keywords:
pharynx, hypopharynx and larynx. Patients with other tumour location were excluded (999, 5.7%). For
Meta-analysis
Systematic review
each tumour location and chemotherapy timing, the logrank-test, stratified by trial, was used to compare
Individual patient data treatments. The hazard ratios of death or relapse were calculated. Interactions between patient or trial
Randomized clinical trials characteristics and chemotherapy effect were studied.
Chemotherapy Results: Individual patient data of 16,192 patients were analysed, with a median follow-up of 5.6 years.
Radiotherapy The benefit of the addition is consistent in all tumour locations, with hazard ratios between 0.87 and 0.88
Head and neck cancer (p-value of interaction = 0.99). Chemotherapy benefit was higher for concomitant administration for all
Oral cavity tumour locations, but the interaction test between chemotherapy timing and treatment effect was only
Oropharynx
significant for oropharyngeal (p < 0.0001) and laryngeal tumours (p = 0.05), and not for oral cavity
Larynx
(p = 0.15) and hypopharyngeal tumours (p = 0.30). The 5-year absolute benefits associated with the con-
Hypopharynx
comitant chemotherapy are 8.9%, 8.1%, 5.4% and 4% for oral cavity, oropharynx, larynx and hypopharynx
tumours, respectively.
Conclusion: The benefit of the addition of chemotherapy to locoregional treatment is consistent in all
tumour locations of HNSCC. The higher benefit of concomitant schedule was demonstrated only for oro-
pharyngeal and laryngeal tumours but this may be only a consequence of a lack of power.
Ó 2011 Elsevier Ireland Ltd. All rights reserved. Radiotherapy and Oncology 100 (2011) 33–40

Head and neck (oral cavity, oropharynx, hypopharynx, and lar-
ynx) squamous cell carcinoma (HNSCC) is a frequent entity with
over 500,000 new cases diagnosed worldwide each year [1]. The
recently updated meta-analysis of chemotherapy in head and neck
cancer (MACH-NC) [2,3], which used the individual patient data of
16,485 patients included in 87 randomized trials performed be-
tween 1965 and 2000, provides level one evidence of a significant
benefit of the addition of chemotherapy in terms of an overall sur-
vival. The hazard ratio of death is 0.88 (p < 0.0001) with an abso-
lute benefit for chemotherapy of 4.5% at 5-years. There is a
⇑ Corresponding author. Address: Meta-analysis Unit, Biostatistics and Epidemi-
ology Department, Institut Gustave Roussy, 114 rue Edouard Vaillant, 94805
Villejuif Cedex, France.
E-mail address: jppignon@igr.fr (J.-P. Pignon).
1 HPV-negative tumours. But it is not known whether treatment ef-
These authors contributed equally to this work.
2
The list of the members of the MACH-NC Collaborative group can be found in Ref.
[3].
significant interaction (p < 0.0001) between chemotherapy timing
(adjuvant, neoadjuvant or concomitant) and treatment. The benefit
is more pronounced for concomitant chemotherapy, with a hazard
ratio of 0.81 (p < 0.0001) and a 5-year absolute benefit of 6.5%. The
benefit of concomitant chemotherapy decreases with increasing
age of the patients (p = 0.003, test for trend). No difference of treat-
ment effect is observed according to tumour site [3].
HNSCC includes four main tumour sites but are usually defined
as one entity. Although these tumour locations share a common
histological type and common major risk factors, they bear specific
characteristics, risk factors, treatment options and prognosis. For
instance, oropharyngeal cancer is specifically associated to human
papilloma virus (HPV) infection, and the prevalence of HPV-related
tumours is increasing [4–6]. HPV related tumours have different
clinicopathologic characteristics and have a better prognosis than
fect varies according to HPV-positivity [6]. The prognosis of HNSCC
depends on tumour site, the highest 5-year survival rates being

0167-8140/$ - see front matter Ó 2011 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.radonc.2011.05.036
34 Head and neck tumour site and chemotherapy
observed for laryngeal carcinomas. Besides, these cancers may not
be treated by the same surgeons: oral cavity cancer can be oper-
ated on by oral/maxillofacial surgeons or by Ear Nose Throat
(ENT) surgeons while larynx/hypopharynx cancers are mainly trea-
ted by ENT surgeons. The therapeutic strategies may differ be-
tween these locations, primary surgery being more frequently
indicated in oral cavity cancers, especially in advanced stages. Lar-
ynx cancers have a different prognosis than hypopharyngeal can-
cers but, due to similarities in treatment strategies, both sites are
usually associated in clinical trials, leading to possible confusion
in the analysis of results, particularly in an organ preservation
setting.
It has therefore been a long-lasting demand by HNSCC special-
ists to have reliable data on treatment efficacy for each tumour
site. A literature-based meta-analysis has been published about
the treatment of oral and oropharyngeal cancers [7], but as most
trials include both oral cavity and oropharyngeal cancer patients,
it is difficult to draw robust conclusions regarding specific tumour
sites. Due to the large number of patients, available covariates and
the updated follow-up, the individual patient data MACH-NC data-
base allows for this analysis.
It was therefore decided to provide a comprehensive analysis of
the MACH-NC database by tumour site. The primary objectives are
to evaluate the benefit of chemotherapy for each tumour location
in terms of overall survival, event-free survival and absolute bene-
fit. The secondary objectives are to investigate interactions be-
tween chemotherapy effect and patient or trial covariates by
tumour site, and to provide comprehensive material for analysis,
discussion and design of future clinical trials for each tumour site.
Material
The material and methods were pre-specified in a protocol
(copy available on request).
Trial search and eligibility criteria
Trial search and eligibility criteria have been described in detail
elsewhere [3]. Briefly, trials were eligible if they had accrued previ-
ously untreated patients with HNSCC and compared locoregional
treatment with locoregional treatment plus chemotherapy. Each
trial had to be randomized in a way that investigators could not
know in advance which treatment the entering patient would re-
ceive (avoiding the potential of allocation bias). Trials were eligible
if accrual was completed between 1965 and 2000, and if all random-
ized patients had undergone a potentially curative locoregional
treatment and had not been treated for another malignancy. Trials
recruiting tumours of the oral cavity, oropharynx, hypopharynx
and larynx were included. Trials including only nasopharyngeal car-
cinomas were excluded. To avoid publication bias, both published
and unpublished trials were included.
Data collection and consistency checking
The data collected for each patient were: age, sex, tumour site,
TNM or stage, performance status, treatment allocated, and date of
randomization. Updated information on survival status and date of
last follow-up were collected. All data were checked for internal
consistency and compared with the trial protocol and published re-
ports and validated with the corresponding trialist.
Description of trials/patients selection process
The meta-analysis included 87 randomized trials (16,485 pa-
tients). References of the trials have been previously published
[2,3]. Because some trials had strata that corresponded to different
locoregional treatments or chemotherapies, and because some tri-
als had 3-arms or a 2 by 2 design, some trial arms were used twice,
such that the number of comparisons in the meta-analysis was 105
and the number of patients 17,493. The flow chart of relevant com-
parisons and patients is summarized in Fig. 1. One trial (n = 36 pa-
tients) was excluded because tumour location was not provided.
Then, patients with tumour locations other than oral cavity, oro-
pharynx, larynx and hypopharynx were excluded (999 patients,
5.7%). Eventually, for statistical reasons, comparisons involving a
group of less than 10 patients were excluded of all analyses (266
patients in 48 strata, 1.5%). Overall 1301 patients (7.4%) were ex-
cluded and 16,192 were included in the present analysis.
Methods
Description of endpoints
The main endpoint was overall survival defined as the time be-
tween randomization and death from any cause. Event-free sur-
vival, defined as the time between randomization and death,
progression or recurrence, was a secondary endpoint. Cancer death
and non-cancer mortality were not studied because there were few
non-cancer deaths in many of the strata, which precluded such a
subgroup analysis.
Analysis
All analyses were carried out on an intention-to-treat basis, i.e.,
all randomized patients were analysed in the allocated treatment
group, irrespective of their actual treatment. Trials were divided
in three groups according to the timing of chemotherapy: adjuvant,
neoadjuvant (also called induction) and concomitant as previously
described [2].
Statistical methods
The median follow-up time was computed according to the re-
verse Kaplan–Meier method [8]. 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 (HR) using a fixed effect model
[9]. Heterogeneity between trials and groups was investigated
using v2 tests [10] and the I2 index [11] that expresses the percent-
age variability of the results related to heterogeneity rather than to
the sampling error. To study the interaction between the treatment
and a covariate, an analysis stratified by trial was performed for
each covariate category (e.g. men/women), and the HRs of the
covariate categories were compared by a test of interaction or
trend as appropriate. Eventually a Cox proportional hazard model
was built to test whether the results remained significant after
adjusting on other patient covariates. For each patient characteris-
tic with a positive interaction or trend test using the first method, a
multivariate model was built including every patient covariate
(age, sex, performance status, stage, chemotherapy) and an inter-
action term between chemotherapy and the positive covariate.
Stratified survival curves were computed for control and experi-
mental groups and used to calculate absolute benefit at 5-years
[12]. The absolute benefit depends on hazard ratio and survival
rate. Sensitivity analyses are similar to those of MACH-NC [2,3].
They were performed as planned in the protocol after the exclusion
of trials conducted before 1980, trials with follow up shorter than
5 years, trials with duplicated control arm or confounded trials, or
after the exclusion of small strata (<40 patients). A random-effects
model was used to take into account unexplained heterogeneity
[13]. All p-values were two sided. A p-value <0.05 was considered
significant, except for heterogeneity analyses where p-values <0.10
 P. Blanchard et al. / Radiotherapy and Oncology 100 (2011) 33–40 35

Available comparisons: n=105 (17 493 patients)

Exclusion of a comparison with no description of tumour location (36 patients)

Available comparisons: n=104 (17 457 patients)

Exclusion of the 999 patients with the following tumour locations:

- nasopharynx (728 patients)
- pharynx, precise location not available (54 patients)
- cervical node (4 patients)
- other/unknown (213 patients)

Available comparisons: n=104 (16 458 patients)

Number of comparisons (patients – percent) excluded due to a Comparisons (patients)

small number of patients (<10 patients per comparison) in the final analysis

Oral cavity: 10 comparisons out of 91 (48 patients - 1,1%) 81 (4 331 patients)

Oropharynx: 8 comparisons out of 90 (41 patients - 0,7%) 82 (5 878 patients)

Larynx: 15 comparisons out of 76 (66 patients - 2%) 61 (3 216 patients)

Hypopharynx: 18 comparisons out of 84 (111 patients - 3,9%) 66 (2 767 patients)

Fig. 1. Flow chart of the comparisons and patients selection process for the purpose of the site-specific analysis.

were considered significant. Statistical analyses were performed
using Statistical Analysis Systems software, version 9.1, for Micro-
soft Windows (SAS Institute, Cary, NC).
Results
Overall results
Description of trials characteristics for each tumour location
Web-table 1 describes trials characteristics according to each
tumour site. Oral cavity tumours are more frequent than orophar-
ynx tumours in trials conducted before 1984 (44% and 25%, respec-
tively), whereas oropharynx tumours are more frequent than oral
cavity cancers in trials conducted after 1991 (42% and 22%, respec-
tively). A detailed description of numbers and percentages of each
tumour location in the included trials is detailed in web-table 2.
Radiotherapy is the cornerstone of local treatment either alone or
after surgery, with rates between 77% and 91% according to tumour
location. Surgery alone is used in 5% of the patients or less, except
for oral cavity cancers where it is used in 11% of the patients. Con-
comitant chemotherapy is the most widely used regimen for all tu-
mour sites. Platinum-based protocols are studied in 47% of oral
cavity tumours, 60% of oropharynx tumours, 47% of larynx tu-
mours, 58% of hypopharynx tumours.
Description of patient’s characteristics for each tumour location
Median follow up is 5.6 years. Details on the characteristics of
the 16,192 patients can be found in web-table 3. Overall, most pa-
tients are men (82%), aged 60 or younger (61%) and in a good phys-
ical condition (performance status 0 or 1 in 87% of patients). Most
patients present with a locally advanced disease, 88% having stage
III or IV disease. Stage I and II are more frequent in larynx cancers
(14%) than in other sites.
Baseline risk according to tumour site
The baseline risk according to tumour site has been estimated as
being the risk of death of stage III-IV patients included in the control
arms of the trials, i.e. without chemotherapy, and after the exclusion
of stage I-II patients. The Kaplan–Meier curves are shown in web-fig-
ure 1. There is a significant survival difference between the four tu-
mour locations (logrank-test p < 0.0001). The highest survival rate is
for laryngeal tumour patients, with a 5-year overall survival esti-
mated at 42%. In this population, the 5-year overall survival rates
of patients with oral cavity, oropharynx and hypopharynx are
respectively 33%, 31% and 27%. To further study the risk of death
according to tumour site, a multivariable model was built including
the following variables: tumour site, tumour stage, performance sta-
tus, age and sex, which included 5755 patients. The results of this
analysis are shown in web-table 4. They show that the better prog-
nosis associated with larynx cancers (HR of death compared to oral
cavity cancers: 0.70; 95% confidence interval (CI): 0.63, 0.78) is inde-
pendent of the other prognostic factors included in the model. Other
independent good prognostic factors are oropharynx cancer (HR:
0.84; 95% CI: 0.76, 0.92), female sex, younger age, early tumour stage
and good performance status. Because performance status was
unavailable in 30% of the patients (see web-table 3), this analysis
was conducted again without the use of performance status as a
covariate. This model, which included 7652 patients (95% of patients
included in the control arms), shows the same results as the previous
model (data not shown) except that oropharynx cancers do not have
a significantly better prognosis anymore, with HR of 0.92 (95% CI:
0.85, 1.01; p = 0.06).
Benefit of chemotherapy according to tumour site and chemotherapy
timing
This section gives an overview of the benefit of adding
chemotherapy according to tumour site and chemotherapy timing.
36 Head and neck tumour site and chemotherapy

No. Deaths / No. Entered

Category LRT+CT LRT O-E Variance Hazard Ratio HR [95% CI]

Oral cavity 1400/2182 1449/2149 -96.2 664.7 0.87 [0.80;0.93]

Oropharynx 1981/2954 2097/2924 -127.4 980.8 0.88 [0.82;0.93]

Larynx 925/1623 949/1593 -60.2 447.1 0.87 [0.80;0.96]

Hypopharynx 958/1380 1001/1387 -58.9 460.6 0.88 [0.80;0.96]

Total 5264/8139 5496/8053 -342.7 2553.1 0.87 [0.84;0.91]

0.25 1.00 4.00

LRT+CT better | LRT better
Test for interaction: p = 0.99 I 2= 19 %
LRT+CT effect: p < 0.0001

Fig. 2. Hazard ratio of death with locoregional treatment plus chemotherapy versus locoregional treatment alone according to tumour site. Abbreviations: CT, chemotherapy;
LRT, locoregional treatment.

Detailed results by tumour site are presented in the next section.
As shown in Fig. 2, it is consistent in all tumour locations (p-value
of interaction = 0.99). Overall, the addition of chemotherapy results
in a reduction of the risk of death of 13% (hazard ratio 0.87, 95% CI:
0.84–0.91) for the patients included in this project. Similar results
are found for event-free-survival (see web-figure 2), with a HR of
progression or death of 0.87 (95% CI 0.84–0.90) in favour of the
addition of chemotherapy.
As previously mentioned, the benefit of chemotherapy varies
according to chemotherapy timing. It is greater for concomitant
administration on the overall population. Table 1 shows the hazard
ratios of death from any cause and the corresponding 5-year abso-
lute benefits associated with chemotherapy according to tumour
location and chemotherapy timing. Table 2 is the corresponding ta-
ble for event-free-survival (EFS). As expected the greater benefit is
seen with the use of concomitant chemotherapy both for OS and
EFS, with HRs around 0.80 and significantly inferior to 1. The interac-
tion test, which tests whether treatment effect on survival varies sig-
nificantly according to chemotherapy timing, is significant for
oropharyngeal and laryngeal tumours (p < 0.0001 and p = 0.05
respectively), but not for oral cavity and hypopharyngeal tumours
(p = 0.15 and 0.30, respectively). The 5-year absolute overall survival
benefits [95% CI] associated with concomitant chemotherapy are
8.9% [4.4, 13.4], 8.1% [4.8, 11.4], 5.4% [0.5, 10.3] and 4.0% [ 1.1,
9.1] for oral cavity, oropharynx, larynx and hypopharynx tumours
respectively (Table 1). The results for EFS are consistent with those
for OS, with 5-year absolute EFS benefits [95% CI] associated with
concomitant chemotherapy are 6.9% [2.8, 11.0], 8.4% [5.1, 11.7],
5.4% [0.7, 10.1] and 3.2% [ 1.7, 8.1] for oral cavity, oropharynx, lar-
ynx and hypopharynx tumours (Table 2). The forest plots showing
the hazard ratios of death (OS) and progression or death (EFS) asso-
ciated with chemotherapy by trial are shown in web-figures 3 and 4.
Detailed results according to tumour location
Oral cavity
4331 patients with oral cavity cancer and 81 comparisons are
included. The HR of death associated with chemotherapy is 0.87
(95% CI: 0.80–0.93), corresponding to an absolute 5-year overall
survival benefit of 5.1% (95% CI: 2.0–8.3), increasing from 31.3%
to 36.4%. The survival curve for oral cavity cancers is shown in
Fig. 3A. According to subset (Table 3) and subgroup analyses

Table 1
Hazard ratios of death and 5-year absolute benefit (overall survival) associated with the use of chemotherapy according to tumour site and chemotherapy timing.

Timing of chemotherapy Test of interaction*

Adjuvant Neoadjuvant Concomitant
Oral cavity HR [95% CI] 0.94 [0.76; 1.17] 0.93 [0.82; 1.05] 0.80 [0.72; 0.89] p = 0.15
5-year abs benefit [CI] +0.4% [ 7.6; 8.4] +2.2% [ 2.9; 7.3] +8.9% [4.4; 13.4]
Oropharynx HR [95% CI] 1.15 [0.92; 1.44] 1.00 [0.90; 1.11] 0.78 [0.72; 0.85] p < 0.0001
5-year abs benefit [CI] 0.4% [ 9.6; 8.8] +1.4% [ 2.9; 5.7] +8.1% [4.8; 11.4]
Larynx HR [95% CI] 1.05 [0.83; 1.33] 1.00 [0.81; 1.23] 0.80 [0.71; 0.90] p = 0.05
5-year abs benefit [CI] +0.1 [ 8.5; 8.7] +3.8% [ 4.6; 12.2] +5.4% [0.5; 10.3]
Hypopharynx HR [95% CI] 1.06 [0.82; 1.38] 0.88 [0.75; 1.02] 0.85 [0.75; 0.96] p = 0.31
5-year abs benefit [CI] 2.3% [ 13.7; ; 9.1] +5.3% [ 0.8; 11.4] +4% [ 1.1; 9.1]

Abbreviations: abs, absolute; CI, 95% confidence interval; HR, hazard ratio.
*
Test of interaction between the HRs.
 P. Blanchard et al. / Radiotherapy and Oncology 100 (2011) 33–40 37

Table 2
Hazard ratios of progression or death and 5-year absolute benefit (event-free survival) associated with the use of chemotherapy according to tumour site and chemotherapy
timing.

Timing of chemotherapy Test of interaction*

Adjuvant Neoadjuvant Concomitant
Oral cavity HR [95% CI] 0.83 [0.69; 1.01] 0.94 [0.84; 1.06] 0.81 [0.73; 0.90] p = 0.16
5-year abs benefit [CI] +5.5% [ 2.1; 13.1] +3.8% [ 1.1; 8.7] +6.9% [2.8; 11.0]
Oropharynx HR [95% CI] 1.09 [0.87; 1.36] 1.05 [0.94; 1.16] 0.74 [0.69; 0.81] p < 0.0001
5-year abs benefit [CI] 0.5% [ 9.5; 8.5] 0.6% [ 4.9; 3.7] +8.4% [5.1; 11.7]
Larynx HR [95% CI] 1.06 [0.85; 1.32] 1.13 [0.92; 1.38] 0.78 [0.70; 0.87] p = 0.002
5-year abs benefit [CI] 1% [ 9.4; 7.4] 1.4% [ 9.6; 6.8] +5.4 [0.7; 10.1]
Hypopharynx HR [95% CI] 0.97 [0.75; 1.25] 0.94 [0.81; 1.09] 0.83 [0.73; 0.93] p = 0.31
5-year abs benefit [CI] +0.5% [ 10.5; 11.5] +3.3% [ 2.4; 9.0] +3.2% [ 1.7; 8.1]

Abbreviations: abs, absolute; CI, 95% confidence interval; HR, hazard ratio.
*
Test of interaction between the HRs.

A. Oral cavity B. Oropharynx

100 Chemotherapy
100 100 100 Chemotherapy
Control
Control
80
71.2 80
Absolute difference [95% CI] 69.8
Absolute difference [95% CI]
68 at 5 years: 5.1% [2.0, 8.3]
Survival (%)

Survival (%)
60
68.4 at 5 years: 5.3% [2.8, 7.8]
51.3
60
50.7
44.4
47.9 40.2 41.1
40 36.4 46 35.8
33.3 40 32.7
38.9 30.4 29.6
34.1 28.1 36 26.6
31.3 30.3 23.7
28.9 26.9 27.4
20 24.5 20 25.4
22.7
HR [95% CI] : 0.87 [0.80 ;0.93] HR [95% CI] : 0.88 [0.82 ;0.93] 20.4

0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Time from randomisation (Years) Time from randomisation (Years)

Number of deaths/person-years: Number of deaths/person-years:

Years 0-2 Years 2-5 Years >=6 Years 0-2 Years 2-5 Years >= 6
LRT+CT 1008/3031 253/2165 139/1620 LRT+CT 1394/4092 434/2729 153/1444
LRT 1070/2846 280/1771 99/1204 LRT 1540/3951 433/2298 124/1255

C. Larynx D. Hypopharynx
100 100 Chemotherapy 100 100
Chemotherapy
Control Control
82.1
80 80.7 80
Absolute difference [95% CI] 71.3
66 at 5 years: 4.5% [0.8, 8.2] Absolute difference [95% CI]
Survival (%)

68 at 5 years: 3.9% [0.2, 7.6]

Survival (%)

60 57.7
62.2 60
51.9
53.2 47 48
47 42.6
39.9 39.5
40 42.5 36.5 40 43.6
33.1
38.3 29.7
35 34.7 26
31.9 24.5
29.8 22.3
25.8
20 20 23.5
21.3 19.6
HR [95% CI] : 0.87 [0.80 ;0.96]
HR [95% CI] : 0.88 [0.80 ;0.96]
0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Time from randomisation (Years) Time from randomisation (Years)

Number of deaths/person-years: Number of deaths/person-years:

Years 0-2 Years 2-5 Years >= 6 Years 0-2 Years 2-5 Years >= 6
LRT+CT 550/2628 253/2126 122/1387 LRT+CT 687/1916 204/1177 67/669
LRT 591/2558 250/1923 108/1146 LRT 764/1866 189/1059 48/564

Fig. 3. Overall survival in trials comparing locoregional treatment plus chemotherapy with locoregional treatment alone for patients with oral cavity (A), oropharynx (B),
larynx (C) and hypopharynx (D) cancers. Abbreviations: CI, confidence interval; CT, chemotherapy; LRT, locoregional treatment.

(Table 4), patients’ age and sex significantly affect chemotherapy Oropharynx
benefit (p-values of interaction: 0.03 and 0.04, respectively), the 5878 patients with oropharyngeal cancer and 82 comparisons
benefit being lower for older patients and men. The Cox multivar- are included. The HR of death associated with chemotherapy is
iate analysis shows that only patients’ sex has a significant inde- 0.88 (95% CI: 0.82–0.93), corresponding to an absolute 5-year over-
pendent interaction with chemotherapy effect (p = 0.009), while all survival benefit of 5.3% (95% CI: 2.8–7.8), increasing from 27.4%
the effect of patients’ age is not significant anymore (p-value of to 32.7% (Fig. 3B). According to subset (Table 3) and subgroup
interaction and trend: 0.27 for both). For event-free survival analyses (Table 4), trials’ year of inclusion, timing of chemotherapy
(web-tables 5 and 6), a similar effect is observed only for sex and type of chemotherapy significantly impact chemotherapy
(p = 0.004; with confirmation in the multivariate analysis, benefit (p-values of interaction: 0.002, <0.0001 and 0.004, respec-
p = 0.0005). tively), suggesting a significantly better effect of chemotherapy in
38 Head and neck tumour site and chemotherapy

Table 3
Hazard ratio of death with locoregional treatment plus chemotherapy versus locoregional treatment alone by trial characteristics for each tumour site.

Overall survival Oral cavity Oropharynx Larynx Hypopharynx

HR [95% CI] p-value HR [95% CI] p-value HR [95% CI] p-value HR [95% CI] p-value
Year of inclusion Before 1984 0.87 [0.78; 0.97] 0.90 (0.80) 1.03 [0.91; 1.16] 0.002 (0.0005) 0.86 [0.75; 0.99] 0.97 (0.80) 0.97 [0.84; 1.12] 0.06 (0.40)
1985–1990 0.88 [0.76; 1.01] 0.88 [0.79; 0.98] 0.88 [0.72; 1.07] 0.75 [0.63; 0.88]
After 1991 0.84 [0.71; 0.99] 0.78 [0.71; 0.86] 0.89 [0.76; 1.04] 0.91 [0.76; 1.08]
Loco regional treatment Standard RT 0.86 [0.77; 0.95] 0.14 0.90 [0.83; 0.98] 0.15 0.82 [0.73; 0.92] 0.30 0.83 [0.72; 0.95] 0.14
Hyperfract. RT 0.61 [0.43; 0.86] 0.73 [0.62; 0.86] 0.76 [0.45; 1.31] 0.85 [0.67; 1.07]
Surgery + RT 0.85 [0.73; 0.98] 0.88 [0.76; 1.03] 0.98 [0.81; 1.19] 1.02 [0.86; 1.21]
Surgery 1.00 [0.73; 1.37] 0.95 [0.34; 2.62] 1.08 [0.56; 2.06] 0.46 [0.23; 0.94]
Others 1.00 [0.81; 1.24] 1.00 [0.81; 1.25] 1.03 [0.76; 1.38] 0.86 [0.62; 1.18]
Timing of chemotherapy Adjuvant 0.94 [0.76; 1.17] 0.15 1.15 [0.92; 1.44] <0.0001 1.05 [0.83; 1.33] 0.05 1.06 [0.82; 1.38] 0.30
Neoadjuvant 0.92 [0.82; 1.05] 1.00 [0.90; 1.11] 1.00 [0.81; 1.23] 0.88 [0.75; 1.02]
Concomitant 0.80 [0.72; 0.89] 0.78 [0.72; 0.85] 0.80 [0.71; 0.90] 0.85 [0.75; 0.96]
Type of chemotherapy Platin + 5 FU 0.90 [0.76; 1.06] 0.90 0.83 [0.75; 0.91] 0.004 0.87 [0.68; 1.10] 0.36 0.84 [0.71; 0.98] 0.14
PolyChemotherapy 0.83 [0.73; 0.96] 0.94 [0.81; 1.08] 0.97 [0.81; 1.17] 1.03 [0.88; 1.21]
Mono CT with platin 0.82 [0.65; 1.04] 0.70 [0.59; 0.84] 0.75 [0.61; 0.93] 0.78 [0.61; 0.99]
Mono CT w/o platin 0.88 [0.78; 0.99] 1.01 [0.89; 1.13] 0.88 [0.76; 1.02] 0.82 [0.68; 0.99]

Abbreviations: CI, confidence interval; CT, chemotherapy; HR, hazard ratio; Hyperfract, Hyperfractionated; 5FU, 5-fluorouracil; p-value indicates the p-value of interaction
(plus p-value of trend when appropriate).

more recent trials, in concomitant chemotherapy trials and in trials
investigating platin or platin + 5-fluorouracil. Patients’ perfor-
mance status and tumour stage are also significantly associated
with chemotherapy benefit (p-values of interaction: 0.004 and
0.02, respectively). The Cox multivariate analysis shows that only
patients’ performance status has a significant independent interac-
tion with chemotherapy effect (p-value 0.0014) with a better effect
of chemotherapy when performance status is 0, while the effect of
tumour stage is not significant anymore (p-value of interaction and
trend: 0.30 and 0.18, respectively). Similar results are observed for
EFS (web-tables 5 and 6), the only significant interaction with pa-
tients’ characteristics being performance status. There is also a sig-
nificant interaction between locoregional treatment and
chemotherapy effect for EFS (p-value: 0.01) which is not found
for OS (p = 0.15), suggesting a better chemotherapy effect in trials
using standard RT or hyperfractionated RT as locoregional
treatment.
Larynx
3216 patients with laryngeal cancer and 61 comparisons are
included. The HR of death associated with chemotherapy is
0.87 (95% CI: 0.80–0.96), corresponding to an absolute 5-year
overall survival benefit of 4.5% (95% CI: 0.8–8.2), increasing from
42.5% to 47.0% (Fig. 3C). Apart from chemotherapy timing, no
subset (Table 3) or subgroup characteristics (Table 4) has a sig-
nificant interaction with chemotherapy benefit. As for OS, there
is no significant interaction between patients’ characteristics
and chemotherapy effect for EFS (web-tables 5 and 6).
Hypopharynx
2767 patients with hypopharyngeal cancer and 66 comparisons
are included. The HR of death associated with chemotherapy is
0.88 (95% CI: 0.80–0.96), corresponding to an absolute 5-year over-
all survival benefit of 3.9% (95% CI: 0.2–7.6), increasing from 25.8%
to 29.7% (Fig. 3D). According to subset (Table 3) and subgroup anal-
yses (Table 4), there is no significant interaction between patients
and trials characteristics and chemotherapy effect, but the trials’
year of inclusion and patients’ age are of borderline significance
(p-value of interaction: 0.06 for trials’ year; p-value of trend for pa-
tients’ age: 0.06). The Cox multivariate analysis shows that the
interaction between patients’ age and chemotherapy effect is of
borderline significance (p-value of trend: 0.05; p-value of interac-
tion: 0.11) suggesting a better effect of chemotherapy in youngest
patients. For EFS, there is no significant interaction with patients’
characteristics, and among trial characteristics, type of locore-
gional treatment and type of chemotherapy have a significant
interaction with chemotherapy benefit (p = 0.01 and 0.05, respec-
tively, see web-tables 5 and 6) suggesting a better effect of stan-
dard/hyperfractionated RT and mono chemotherapy with platin.
Sensitivity analyses
For each tumour location, excluding trials performed before
1980, or confounded, or with strata smaller than 40 patients or
with a follow-up shorter than 5-years or with a duplicated control
arm does not change the results (sensitivity analysis, see web-table
7). The only differences are that the effect of chemotherapy timing
became non significant for laryngeal tumours after the exclusion of
small subgroups, old trials and confounded trials and significant for
hypopharyngeal tumours after the exclusion trials with follow-up
65 years.
Heterogeneity analysis
Heterogeneity is mainly located in oral cavity cancers
(p = 0.0002, see web-table 7A). Indeed the test for heterogeneity
is of lower significance for oropharynx cancers (p = 0.08, web-table
7B) and not significant in the other tumour sites (respectively
p = 0.75 and p = 0.30 for larynx and hypopharynx cancers, see
web-table 7C and D). Regarding the oral cavity subgroup, this het-
erogeneity is located in the concomitant chemotherapy trials
group (p-value of heterogeneity of 0.0001, 0.81 and 0.29 for the
concomitant, adjuvant and neoadjuvant groups respectively).
Exclusion of oldest trials, smallest trials or trials with a follow-up
inferior to 5 years as part of the sensitivity analysis lowered this
heterogeneity (p = 0.04, 0.002 and 0.01 respectively). The use of a
random-effects model did not change neither the estimates of
the benefit of chemotherapy nor the absence of interaction be-
tween chemotherapy timing and benefit.
Discussion
This analysis provides the highest level of evidence regarding
the benefit of chemotherapy for the different tumour locations of
locally advanced head and neck cancers. The addition of
chemotherapy to local treatment confers a survival benefit, which
is consistent in all four tumour locations.
The MACH-NC database consists of HNSCC patients included in
randomized trials of chemotherapy over a 35 years period.
Although it is known that, due to inclusion criteria, clinical trial pa-
 P. Blanchard et al. / Radiotherapy and Oncology 100 (2011) 33–40 39

Table 4
Hazard ratio of death with locoregional treatment plus chemotherapy versus locoregional treatment alone by patient characteristics for each tumour site.

Overall survival Oral cavity Oropharynx Larynx Hypopharynx

HR [95% CI] p-value HR [95% CI] p-value HR [95% CI] p-value HR [95% CI] p-value
Age 650 0.87 [0.75; 1.01] 0.03* (0.16) 0.86 [0.76; 0.98] 0.14 (0.14) 0.76 [0.58; 0.98] 0.54 (0.39) 0.76 [0.61; 0.95] 0.18 (0.06)
51–60 0.76 [0.67; 0.87] 0.83 [0.75; 0.93] 0.89 [0.76; 1.04] 0.86 [0.73; 1.01]
61+ 0.99 [0.86; 1.13] 0.97 [0.87; 1.08] 0.89 [0.77; 1.02] 0.98 [0.84; 1.14]
Sex Male 0.91 [0.84; 0.99] 0.04** 0.89 [0.83; 0.96] 0.50 0.86 [0.78; 0.95] 0.78 0.86 [0.78; 0.95] 0.29
Female 0.73 [0.61; 0.88] 0.83 [0.69; 1.00] 0.90 [0.66; 1.23] 1.04 [0.74; 1.46]
Performance status 0 0.92 [0.79; 1.07] 0.60 0.73 [0.64; 0.82] 0.004*** 0.87 [0.74; 1.01] 0.64 0.84 [0.70; 1.00] 0.63
1+ 0.87 [0.77; 0.98] 0.91 [0.83; 0.99] 0.82 [0.70; 0.97] 0.79 [0.68; 0.92]
Stage I, II 0.90 [0.66; 1.24] 0.60 (0.60) 0.75 [0.56; 1.00] 0.02**** (0.20) 0.89 [0.63; 1.24] 0.98 (0.93) 1.01 [0.60; 1.70] 0.52 (0.26)
III 0.80 [0.68; 0.93] 1.01 [0.88; 1.14] 0.85 [0.72; 1.01] 0.94 [0.77; 1.13]
IV 0.87 [0.79; 0.96] 0.83 [0.77; 0.90] 0.86 [0.76; 0.97] 0.84 [0.75; 0.94]

Abbreviations: CI, confidence interval; HR, hazard ratio; p-value indicates the p-value of interaction (plus p-value of trend when appropriate).
*
Adjusted p-value = 0.27 (Cox model).
**
Adjusted p-value = 0.009 (Cox model).
***
Adjusted p-value = 0.0014 (Cox model).
****
Adjusted p-value = 0.30 (Cox model).

tients are somewhat different of patients of the ‘‘real life’’ [14], it is
worth noting that MACH-NC patients are mostly male in their sixth
or seventh decade, suffering from locally advanced cancers. Apart
from the good performance status (48% of patients are PS 0), and
a probably larger proportion of advanced stages, this large popula-
tion is similar to the general population of HNSCC patients. We
provide baseline risk curves after exclusion of stage I and II pa-
tients in order to use a homogeneous population and limit the dif-
ferences in outcome that might be related to other factors than
tumour location. The majority of stage I and II patients are in the
larynx cancer group of patients, which is consistent with the clin-
ical routine practice and contributes to the fact that larynx cancer
is of better prognosis than the other sites. Oral cavity cancers are
more frequent in the first period of trial inclusion time than
oropharynx cancers but this tendency has inverted with time. This
may be related with the increase of human papilloma virus posi-
tive oropharynx cancers incidence [4–6], although lots of confusing
factors may have played a role in the selection of these patients.
Overall the results presented here are applicable to the general
population.
Larynx cancers have a better prognosis than other HNSCC. This
is partly due to the fact that larynx cancers are early symptomatic,
therefore many larynx cancers are diagnosed at an early stage,
added to the fact that early glottic cancers are much less lympho-
philic than other HNSCC. This large database confirms that the
prognosis of larynx cancer remains better than other HNSCC, even
after adjustment on other major prognostic factors and exclusion
of early stages.
The addition of chemotherapy to locoregional treatment results
in a reduction of the risk of death of 13%, consistent in all tumour
sites. The 5-year absolute benefits associated with concomitant
chemotherapy are around 8% for oral cavity and oropharynx can-
cers and around 5% for hypopharynx and larynx cancers. The inter-
action between chemotherapy timing and survival is significant
only for oropharynx and larynx cancers, but the same trend to-
wards a better efficacy of concomitant chemotherapy can be ob-
served in oral cavity and hypopharynx cancers. A lack of power
probably explains the non-significance of the interaction test in
these sites. It should also be of note that trials using taxane, cisplat-
inum and 5-fluorouracil (TPF) as neoadjuvant chemotherapy are
not included in this database so the results cannot be applied in
this setting [15,16].
Lack of power and the risk of false positive are the major limi-
tations of this study. Indeed, whereas the power of the meta-anal-
ysis to show differences on the entire population is very high, it is
much lower when the same factors are analysed separately for
each tumour location. Therefore negative results, such as the ab-
sence of interaction between chemotherapy timing and outcome,
might be only related to a lack of power. In the absence of signifi-
cant differences, the overall estimate may be more accurate than
the one based on subgroup analyses. On the other hand, repeating
subgroup analyses increases the risk of false positive results. This
has been limited by the systematic use of a confirmatory multivar-
iate Cox model for each significant interaction between patients’
characteristics and outcome. Only the significant interaction in
both models and on both endpoints should be considered as true
positive. The last potential limitation of this study is related to
the exclusion of small patient strata. This was decided to reduce
statistical heterogeneity. But as shown in web-table 8, it did not
change the HR related to the addition of chemotherapy nor the
heterogeneity.
In the oropharynx cancers group, there was a significant
interaction between chemotherapy timing, type of chemotherapy,
performance status and treatment effect. The results on type and
timing of chemotherapy are consistent with the overall results of
MACH-NC. This group is the largest group of the database, with
5919 patients. There is thus more power to detect statistical inter-
actions. There was also a significant interaction between the year
of inclusion and treatment effect, which was not observed in the
other tumour locations. These results are partly explained by the
interaction with chemotherapy type and timing, as the most recent
trials are concomitant trials using platinum salts. Although there is
no strong evidence showing that HPV positive tumours gain a
higher benefit from the addition of chemotherapy than HPV nega-
tive tumours [6], the rising incidence of HPV positive oropharyn-
geal cancers might have played a role in this improvement with
time of the treatment effect and should be further studied in pro-
spective clinical trials.
In the hypopharynx cancer group, the decrease of chemother-
apy benefit in older patients was of borderline significance. The
same result, although non significant, was observed for all tumour
sites, the oldest age category always having the hazard ratios close
to 1. This trend is significant on the overall database and was sig-
nificant in the meta-analysis of fractionation in head and neck can-
cer [17]. There is no reason to believe that this interaction varies
according to the tumour site. A lack of power is the most likely
explanation for the non-significance of this trend in the other
groups.
Event-free survival analyses are consistent with OS analyses.
There are some minor differences with a few significant interac-
tions between trial covariates and treatment effect in some site
groups. This is consistent with the fact that EFS is a good surrogate
40 Head and neck tumour site and chemotherapy
marker for overall survival in head and neck cancer patients trea-
ted with radiotherapy and chemotherapy [18]. As a comment, it
would be interesting to see if this consistent effect on each head
and neck subsite remains true for other treatment approaches,
such as the use of hypoxic modifications in HNSCC [19] or altered
modification radiotherapy. In a recent negative trial of accelerated
radiotherapy in HNSCC [20], the benefit was limited to the sub-
group of patients with oral cavity cancer. Similarly and despite
the fact that recent papers on tumour control after IMRT focus
on the pattern of relapse according to the tumour subsite [21], this
area should be further investigated.
To conclude, we provide an extensive analysis of the prognosis
and the benefit of chemotherapy for each tumour location of head
and neck cancers in terms of overall survival, event-free survival
and absolute benefit in the largest worldwide available prospective
database. This approach provides survival curves and clinical data
that could be used as reference data for observational studies or
clinical trials, as this unique database covers most published and
unpublished randomized trials worldwide over a 35 year period.
Overall, the benefit of the addition of chemotherapy to locoregional
treatment is consistent in all tumour locations of HNSCC, is pre-
dominant in the concomitant schedule and is lower in the oldest
patients.
Role of the funding source
The sponsors of this study had no role in the study design, data
collection, data analysis, data interpretation, or the writing of the
report.
Acknowledgments
The chief acknowledgement is to the trialists who agreed to
share their data. We also thank the following institutions for fund-
ing the investigators meeting or the meta-analysis project: Associ-
ation pour la Recherche sur le Cancer (ARC No. 2015), Institut
Gustave-Roussy, Ligue Nationale Contre le Cancer, Programme
Hospitalier de Recherche Clinique (No. IDF 95009), Sanofi-Aventis.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.radonc.2011.05.036.
References
[1] Ferlay J, Shin HR, Bray F, Forman D, Mathers C, Parkin DM. GLOBOCAN 2008.
Cancer Incidence and Mortality Worldwide: IARC CancerBase No. 10 [Internet].
Lyon, France: International Agency for Research on Cancer; 2010. Available
from: http://globocan.iarc.fr.
[2] Pignon JP, Bourhis J, Domenge C, Designe L. Chemotherapy added to
locoregional treatment for head and neck squamous-cell carcinoma: three
meta-analyses of updated individual data. MACH-NC Collaborative Group.
Meta-Analysis of Chemotherapy on Head and Neck Cancer. Lancet
2000;355:949–55.
[3] Pignon JP, le Maitre A, Maillard E, Bourhis J. Meta-analysis of chemotherapy in
head and neck cancer (MACH-NC): an update on 93 randomized trials and
17,346 patients. Radiother Oncol 2009;92:4–14.
[4] Bleyer A. Cancer of the oral cavity and pharynx in young females: increasing
incidence, role of human papilloma virus, and lack of survival improvement.
Semin Oncol 2009;36:451–9.
[5] Hammarstedt L, Dahlstrand H, Lindquist D, et al. The incidence of tonsillar
cancer in Sweden is increasing. Acta Otolaryngol 2007;127:988–92.
[6] Lassen P. The role of Human papillomavirus in head and neck cancer and the
impact on radiotherapy outcome. Radiother Oncol 2010;95:371–80.
[7] Furness S, Glenny AM, Worthington HV, et al. Interventions for the treatment
of oral cavity and oropharyngeal cancer: chemotherapy. Cochrane Database
Syst Rev 2011;4:CD006386.
[8] Schemper M, Smith TL. A note on quantifying follow-up in studies of failure
time. Control Clin Trials 1996;17:343–6.
[9] Yusuf S, Peto R, Lewis J, Collins R, Sleight P. Beta blockade during and after
myocardial infarction: an overview of the randomized trials. Prog Cardiovasc
Dis 1985;27:335–71.
[10] Non-mall Cell Lung Cancer Collaborative Group. Chemotherapy in non-small
cell lung cancer: a meta-analysis using updated data on individual patients
from 52 randomized clinical trials. BMJ 1995;311:899–909.
[11] Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat
Med 2002;21:1539–58.
[12] Early Breast Cancer Trialists’ Collaborative Group. Systemic treatment of early
breast cancer by hormonal, cytotoxic, or immune therapy. 133 randomized
trials involving 31,000 recurrences and 24,000 deaths among 75,000 women.
Lancet 1992;339:1–15.
[13] Michiels S, Baujat B, Mahe C, Sargent DJ, Pignon JP. Random effects survival
models gave a better understanding of heterogeneity in individual patient data
meta-analyses. J Clin Epidemiol 2005;58:238–45.
[14] Talarico L, Chen G, Pazdur R. Enrollment of elderly patients in clinical trials for
cancer drug registration: a 7-year experience by the US Food and Drug
Administration. J Clin Oncol 2004;22:4626–31.
[15] Posner MR, Hershock DM, Blajman CR, et al. Cisplatin and fluorouracil alone or
with docetaxel in head and neck cancer. N Engl J Med 2007;357:1705–15.
[16] Vermorken JB, Remenar E, van Herpen C, et al. Cisplatin, fluorouracil, and
docetaxel in unresectable head and neck cancer. N Engl J Med
2007;357:1695–704.
[17] Bourhis J, Overgaard J, Audry H, et al. Hyperfractionated or accelerated
radiotherapy in head and neck cancer: a meta-analysis. Lancet
2006;368:843–54.
[18] Michiels S, Le Maitre A, Buyse M, et al. Surrogate endpoints for overall survival
in locally advanced head and neck cancer: meta-analyses of individual patient
data. Lancet Oncol 2009;10:341–50.
[19] Overgaard J. Hypoxic modification of radiotherapy in squamous cell carcinoma
of the head and neck – a systematic review and meta-analysis. Radiother
Oncol 2011;100:22–32.
[20] Zackrisson B, Nilsson P, Kjellén E, et al. Two-year results from a Swedish study
on conventional versus accelerated radiotherapy in head and neck squamous
cell carcinoma – the ARTSCAN study. Radiother Oncol. 2011 Feb 3. [Epub
ahead of print].
[21] Loimu V, Collan J, Vaalavirta L, et al. Patterns of relapse following definitive
treatment of head and neck squamous cell cancer by intensity modulated
radiotherapy and weekly cisplatin. Radiother Oncol 2011;98:34–7.