48 / 220
69.82;
Figure 7
). The AUC was 0.93 (standard error 0.02)
with a Q
*
index of 0.86 (standard error 0.02).
Meta-regression Analysis
On univariate meta-regression analysis, sample size, QUADAS
score, imaging type, timing of posttreatment scan, and method
of image interpretation did not affect the diagnostic odds
ratio for detection of local, nodal, or overall tumor recur-
rence (all
P
values
.
.05). There were insufficient data to
assess the impact of the initial tumor site and clinical pre-
sentation at recurrence on test accuracy. Because of the
limited number of studies reporting on the detection of dis-
tant recurrences, the effect of these factors on test accuracy
could also not be assessed.
Subgroup Analysis
Subgroup analysis could not be performed for studies evalu-
ating the detection of distant metastases as there were insuf-
ficient studies.
PET versus PET/CT.
PET/CT was found to be more specific
than PET alone in the detection of residual/recurrent disease
at the primary site (
P
\
.001). No significant difference in
sensitivity was noted between the 2 modalities for local
recurrence (
P
= .07). There was no statistical difference in
the sensitivity or specificity between studies using PET to
detect residual/recurrent disease at neck sites compared with
those using PET/CT (
Table 2
).
Visual vs semiquantitative analysis.
No statistical difference
was found in the sensitivity or specificity between visual
and semiquantitative analysis of scans for the detection of
disease at the primary site or in the neck (
Table 2
).
Timing before 12 weeks versus after 12 weeks.
Studies that
had scans performed both before and after the 12-week
cutoff were excluded from the subgroup analysis. The speci-
ficity was significantly higher for scans performed more
than 12 weeks after treatment for both local and nodal
recurrence (
P
= .009 and
P
= .0043, respectively). There
was no significant difference found in the sensitivity of
scans (
Table 2
).
Discussion
The aim of our meta-analysis was to evaluate the diagnostic
accuracy of PET and PET/CT for the detection of residual
and/or recurrent disease in the post-(chemo)radiotherapy set-
ting. We found that PET and PET/CT were highly accurate
in the detection of residual and/or recurrent disease at local,
nodal, and distant sites, although the timing of the scan did
have an impact on the accuracy of such scans. PET/CT was
more specific than PET alone in the detection of disease at
the primary site. However, no difference was found between
scans that were interpreted visually compared with those ana-
lyzed semiquantitatively using standard uptake values.
While there have been previous meta-analyses summariz-
ing the diagnostic accuracy of PET and PET/CT in the
detection of recurrences at locoregional
38,39
and distant
40
sites, these reviews have included retrospective as well as
prospective studies, and this may overestimate the diagnos-
tic test accuracy by introducing bias. Moreover, the meta-
analysis by Gao et al
40
included patients with head and
neck cancers, not specifically SCCs, while the study by
Isles et al
39
included data on dual-head gamma detection
systems that have inferior resolution compared with dedi-
cated full-ring PET scanners.
Figure 6.
Forest plots of sensitivity and specificity for positron emission tomography and positron emission tomography/computed tomo-
graphy in the diagnosis of distant metastases in recurrent head and neck squamous cell carcinomas.
Cheung et al
28