57 / 220
2.3. Statistical analyses
Groups were compared in terms of categorical variables
by using the Fisher’s exact test. PET/CT and CWU were
compared in terms of the sensitivity and accuracy with
which they staged HNSCC by using the McNemar test.
The Kaplan
e
Meier method was used to estimate overall
survival (OS) and index cancer progression (progres-
sion-free survival [PFS]). OS and PFS were defined as
the time between the first day of treatment to the date of
death or progression, respectively, or to the last clinical
follow-up. Disease progression was defined as the
appearance of new lesions or enlargement of the initial
primary tumour and/or development of metastatic dis-
ease
[22]
. The log-rank test was used to compare sur-
vival rates according to stage and management impact.
The Cox proportional hazards model was used to
identify the prognostic variables for univariate and
multivariate predictions of PFS and OS. The tests were
based on the likelihood ratio statistic, and the estimated
hazard ratio (HR) and 95% confidence intervals (CIs)
were calculated. All statistical analyses were two sided
and were performed by using SPSS software, version
22.0.
P
<
0.05 was deemed to indicate statistical
significance.
3. Results
The patient clinical characteristics are presented in
Table 1
. PET/CT changed the CWU-based TNM clas-
sification of 83 lesions in 79 (31.9%) patients; in two
patients, both T and N changed. In the remaining two
patients, both N and M changed. In the remaining 169
(68.1%) patients, PET/CT and CWU findings presented
identical TNM classification. Of the 79 patients with
discordant TNM classification, histopathology was
available in 68 (86.0%). In another two and one patient,
the TNM classification was confirmed by subsequent
imaging and clinical follow-up, respectively. It was not
possible to definitively confirm the stage in the remain-
ing eight patients. These cases were not included when
comparing the PET/CT stage and CWU stage in terms
of diagnostic accuracy.
The discrepancies of T classification were identified in
24 (30.3%) patients. The extent of the primary tumour
was not confirmed in two patients because ICT was
performed. There were no false-positive PET/CT results
for the detection of primary tumour. However, PET/CT
failed to detect 15 of the CWU-staged T1 tumours (15 of
101, 14.8%). The discrepancies of N classification were
identified in 55 (69.6%) patients (
Supplementary Table
S1
). The discordant nodal stage was confirmed by his-
topathology in 46 and serial imaging in 1. PET/CT
classified the N classification more accurately and
sensitively than CWU (both
P
<
0.05,
Supplementary
Table S2
). The discrepancies of M classification were
identified in four (5.0%) patients. PET/CT downstaged
two of four patients, correctly in all cases, and upstaged
one patient correctly. The remaining one was not
assessable because of therapeutic interventions for un-
confirmed site of disease visualised only on imaging.
In terms of detecting synchronous SPCs, CWU and
PET/CT differed in 21 (11%) patients. CWU detected
SPCs in 12 of these 21 patients: these SPCs were in the
oesophagus (
n
Z
4), stomach (
n
Z
2), thyroid (
n
Z
2),
or lung (
n
Z
4). PET/CT accurately excluded the SPC in
the lung in three of the latter four cases. However, PET/
CT failed to detect the four cases of CWU-detected SPC
in the oesophagus: this was inaccurate. PET/CT accu-
rately detected the remaining one in the lung, the two in
the stomach, and the two in the thyroid. Additionally,
PET/CT, but not CWU, detected SPCs in the remaining
nine patients: the SPCs were in the thyroid (
n
Z
3), lung
(
n
Z
2), colon (
n
Z
1), breast (
n
Z
1), palatine tonsil
(
n
Z
1), and epiglottis (
n
Z
1).
Overall, PET/CT stage and CWU stage were discor-
dant in 79 patients (31.9%), for whom a validation was
available in 71 patients. PET/CT staging was signifi-
cantly more sensitive and accurate than CWU staging
(both
P
<
0.001;
Table 2
). Considering the whole pop-
ulation of the study, we hypothesise that the stages
would be correct for patients with identical PET/CT and
CWU
e
TNM stages because changes of management
were not expected in these patients. The overall accuracy
of PET/CT staging was significantly higher than those of
CWU staging (87.1% versus 82.0%;
P
<
0.001).
Patients whose CWU-determined stage was upstaged
by PET/CT staging had a significantly worse PFS and
OS than those whose clinical stage did not change (both
P
<
0.05) (
Fig. 1
).
3.1. Primary outcome
Overall, the PET/CT staging led to management
changes in 39 (15.7%) of the 248 patients. PET/CT had a
significantly higher impact on the CWU-determined
management plan in patients who were CWU staged
as III
e
IV than in those who were CWU staged as I
e
II
(21.4% versus 9.8%,
P
Z
0.014).
In 12 patients (4.8%), PET-CT had a high impact on
the CWU-determined management plan. In most cases,
this was because PET/CT detected distant metastasis
and SPCs (details are given in
Supplementary Table S3
).
All these lesions were histologically confirmed. PET/CT
correctly changed the management of these patients. In
27 (10.9%) patients, PET/CT had a moderate impact on
the management plan. This was mainly because PET-CT
upstaged the nodal stage. This led to modification of the
radiation field and/or dose (
n
Z
9) and surgical extent
(
n
Z
18). The actual disease stage could be validated in
24 patients: PET/CT correctly changed the management
in 19 of these patients. PET/CT had a low impact in 206
(83%) patients, predominantly because it concurred with
I.S. Ryu et al. / European Journal of Cancer 63 (2016) 88
e
96
37