101 / 220
conducted using R version 2.15.1 (R Foundation for Statistical
Computing, Vienna, Austria) and the metafor package (16).
Results
Figure 1 shows the flowchart of studies retrieved and ex-
cluded. Of the 1822 titles initially identified from the database
search, 41 full-length articles were assessed for inclusion, of
which 27 were excluded and 14 studies were determined to be
eligible and were included in this systematic review (7,17–29).
Appendix Table A1 lists these 27 articles (6,8,30–54) and the
reason for their exclusion. No additional study was found
from our search of the three bibliographies in previous meta-
analyses (9–11). One study (8) was excluded, as it analyzed a
subset of study subjects that were later recruited in a multi-
center cohort study (25).
Baseline characteristics
Table 1 shows a comparison of the baseline characteristics
between the 14 eligible studies. There was no randomized
trial. Thirteen studies were retrospective, while one was
prospective. Of the 3331 patients included, 1592 (47.8%) un-
derwent TT
+
pCND (group A), while 1739 (52.2%) under-
went TT only (group B). In terms of preoperative nodal
assessment, ultrasonography (US) was used as the standard
imaging modality in all studies, but only two studies specif-
ically mentioned that both bilateral central and lateral neck
compartments were examined (25,27).
In terms of selection for pCND, seven studies were based
on individual surgeon’s preference (7,18,22,24,25,27,28),
while four studies did not specify their method of selection
(17,19,23,26). Three studies used historical controls (TT alone)
for comparison (20,21,29). Only 11 of 14 studies statistically
compared age, sex ratio, tumor size, extrathyroidal extension,
and tumor multifocality between the two groups (7,17,18,21–
26,28,29). Of these, two studies found age to be significantly
older in group B (21,25), and three studies found tumor size to
be significantly different (7,21,23). Two studies found tumor
size to be significantly larger in group A (7,21), while one
study found tumor size to be significantly smaller in group A
(23). Three of nine studies found the rate of extrathyroidal
extension to be significantly higher in group A (7,22,24), and
two of nine studies found the rate of tumor multifocality to be
significantly higher in group A (22,26). Bilateral pCND was
performed in eight studies (17,18,22–24,26,28,29), while the
other six studies performed either unilateral or a combination
of unilateral and bilateral pCND (7,19,20,21,25,27). Among the
eight studies reporting bilateral pCND (17,18,22–24,26,28,29),
the mean number of central lymph nodes harvested ranged
between 5.6 and 9.6, while the one study reporting unilateral
pCND harvested a median of five (7). The incidence of central
LNM in group A ranged from 23.5% to 82.4%, while in group B
it ranged from 0.9% to 9.7% with 9 of 14 studies not reporting
the incidence of central LNM in group B.
Surgical outcomes
Table 2 shows a comparison of outcomes between the two
groups. Only 9 of the 14 studies reported whether RAI abla-
tion was given after surgery (7,18,21–24,26,27,29). Their dose
ranged from 2.78 to 5.55 GBq. One study empirically gave the
same dose of RAI, irrespective of the extent of LNM (7). The
mean frequency of postoperative RAI ablation in groups A
and B were 746/1041 (71.7%) and 498/937 (53.1%). Group A
was significantly more likely to receive RAI ablation than
group B (OR
=
2.60 [CI
=
2.12–3.18]). This was expected be-
cause of the higher incidence of central LNM (or N1a) in
group A leading to tumor group upstaging in patients older
than 45 years (28). Only 1 of 14 studies compared operating
time between the two groups and found group B to have a
significantly shorter operating time than group A (28).
Figure 2 shows the forest plot for temporary hypocalcemia.
Of the 14 studies, 11 studies compared temporary postoper-
ative hypocalcemia between the two groups, while 10 studies
compared permanent postoperative hypocalcemia in the two
groups. In eight studies, permanent hypocalcemia was de-
fined as persistent hypocalcemia and/or need for calcium
supplements for more than six months (7,20,23–28), while two
studies defined it as more than 12 months (18,29). If one as-
sumed all studies utilized a similar definition for temporary
and permanent hypocalcemia, the overall temporary hypo-
calcaemia rate in group Awas significantly higher than that in
B (336/1294 (26.0%) and 144/1330 (10.8%), respectively;
OR
=
2.56 [CI
=
2.04–3.21]) while the overall permanent hypo-
calcaemia was also similar between the group A and B (25/
1254 (2.0%) and 15/1257 (1.2%), respectively; OR
=
1.74
[CI
=
0.87–3.50]).
FIG. 1.
Flow diagram for study selection.
PROPHYLACTIC CND DID NOT SIGNIFICANTLY LOWER LRR
81