Can we predict the first variceal bleeding in the individual cirrhosis and esopna~ui V~IICGJi
I” the pas
fewyears. the
ability
10
predict the
fm, van-
bleeding has been somewhat improved by the analysis of
cd
iarge groups of patients with cirrhosir
and esophsgent van-
ces, and by the USC of sophisticated statistical technology. The NIEC
index, proposed by an Italian multicenter study
group in 1988 (I), IS presently used m manygastroenterologicat units to assess the nsk of first bleeding in such pntients. The mdex 1s easily calculated from currcnrly available data (size of “arices, Child’s clarification.
severity of red color
signs). However.
the clinical usefulness of the NlEC
index in
terrm of sensitivity and specificity, of its ability to detect who wdl bleed and who will not, has not been clearly addressed. Using the data from the NlEC
patient with
.2
20
‘V
:$ 0.8 i
25
,j
/
c’
/’
/’
04
30
,/+l
I
0.4
,
55
I
/
I’
,/’
I
study, we calculated the
RDC curve of the plediction of bleeding during follow-up in relation to the NIEC index, and compared it with the ROC curve ohtamed when patients were only divided according 10 the she of varices (Fig. I). Comparison was performed by measuring the areas under the curve (AK)
and their stan-
dard errors (SE.) according to Hanley and McNeil (2,3). The NIEC index was more effnenr than the standard classification of varicea, sire (AUC = 0.711; SE. = O.lu‘l; vs. AUC
= 0.630; S.E.
= 0.0X),
and the difference war proba-
bly statistically significant. In fact, since the average mrrelation mefticient
between areas (not available from Dublished
data) probably ranged 0.3-01,
the L statistics ‘probably
raneed 1.95-2.30. with associated o values of 0.026-0.011 (on; tail). However, the ROC c;rve obtamed from the NIEC
index appeared somewhat far from the upper leh mr-
ner. whxeh 15the ideal condition. Adequate specificity (88%) was obtamed only at low sensitivity stance. if we chose a NIEC conversely a good sensithty
VBIUCS (38%.
for in-
index of 35 as cut-off point), and (78%)
was associated with a
substantial lack of specificity (51%, for instance, if we chose 25 ascut-off point). ‘Therefore, it is our conclusion that better prognostic indicat~fs of the first variceal bleeding in patients with cirrhosis are still needed.