ABSTRACT
training of paediatric endoscopists may improve the agreement in grading
Objectives: Data regarding agreement on endoscopic features of oesopha-
severity of varices in this setting.
geal varices in children with portal hypertension (PH) are scant. The aim of
this study was to evaluate endoscopic visualisation and classification of Key Words: children, endoscopy, interobserver agreement, oesophageal
oesophageal varices in children by several European clinicians, to build a varices, portal hypertension
rational basis for future multicentre trials.
Methods: Endoscopic pictures of the distal oesophagus of 100 children with (JPGN 2015;61: 176181)
a clinical diagnosis of PH were distributed to 10 endoscopists. Observers
were requested to classify variceal size according to a 3-degree scale (small,
medium, and large, class A), a 2-degree scale (small and large, class B), and
to recognise red wales (presence or absence, class Red). Overall agreement What Is Known
was considered fair if Fleiss and Cohen k test was 0.30, good if 0.40,
excellent if 0.60, and perfect if 0.80. The natural history and response to treatment for
Results: Agreement between observers was fair with class A (k 0.34) and
oesophageal varices in children is not completely
class B (k 0.38), and good with class Red (k 0.49). The agreement was
understood.
good on presence versus absence of varices (class A 0.53, class B 0.48). Prospective studies need to be multicentric to
The agreement among the observers was good in class A when endoscopic
achieve sufficient statistical power.
features of severe PH (medium and large sizes, red marks) were grouped and There are few data regarding agreement on classifi-
compared with mild features (absent and small varices) (k 0.58).
cation of varices in children.
Conclusions: Experts working in different centres show a fairly good
agreement on endoscopic features of PH in children, although a better
What Is New
paediatric PH will need to be multicentric if they are to have As a speculation, we then calculated the different rate of
sufficient power (6), and good agreement between endoscopists treatment that would be applied if different classifications and
in different tertiary care centres will be fundamental to such different criteria were used to indicate prophylaxis.
studies (7). Therefore, our aim was to investigate interobserver The observers received the different classifications named on
agreement amongst endoscopists working in different settings in a PowerPoint presentation without providing visual examples for
assessing the presence and severity of oesophageal varices in the scoring scale, to measure and describe the agreement without
children with PH. affecting the usual observers practice.
80
1. Classification A (class A) considered 3 degrees of variceal size.
Grade 0: no oesophageal varices; grade 1: small and 70
nontortuous oesophageal varices; grade 2: tortuous oesophageal
varices but occupying less than one-third of the distal 60
oesophagus radius; grade 3: large and tortuous oesophageal
Percentage
50 No varices
varices covering more than one-third of the distal oesophagus Small
radius (8,9). 40 Medium
www.jpgn.org 177
Copyright 2015 by ESPGHAN and NASPGHAN. Unauthorized reproduction of this article is prohibited.
DAntiga et al JPGN Volume 61, Number 2, August 2015
Classification B (Class B)
Grade 2: medium
The overall agreement with the reference observer was 0.38
(range 0.260.52) (fair agreement, Cohen k test) (Fig. 4); the group
agreement with the observer was good for the presence of
varices (k 0.48, range 0.300.63) and good for the presence
of large varices (k 0.42, range 0.240.54) (Table 1).
Small Varices With Red Marks and Medium and Small Varices With Red Marks and Large Varices in
Large Varices in Class A Class B
Considering the guidelines proposed in adults to indicate Considering the guidelines proposed in adults to indicate
prophylaxis, including small varices with red marks, as well as prophylaxis, including small varices with red marks, as well as large
medium and large varices, the mean agreement between the group varices, the mean agreement between the group and the reference
and the reference observer in class A was good with Cohen test observer in class B was good with Cohen test (k 0.46, range
178 www.jpgn.org
Copyright 2015 by ESPGHAN and NASPGHAN. Unauthorized reproduction of this article is prohibited.
JPGN Volume 61, Number 2, August 2015 Endoscopic Classification, Oesophageal Varices
TABLE 1. Fleiss and Cohen coefficient of agreement within the group for the studied classifications
Fleiss (95% CI) 0.34 (0.330.36) 0.38 (0.360.40) 0.48 (0.450.51) 0.48 (0.450.50) 0.45 (0.420.48) 0.57 (0.540.60) 0.46 (0.440.49) 0.53 (0.500.56)
Cohen (range) 0.34 (0.230.45) 0.38 (0.260.52) 0.49 (0.400.59) 0.53 (0.300.71) 0.48 (0.300.63) 0.60 (0.490.68) 0.46 (0.400.61) 0.58 (0.440.72)
Fleiss k test refers to the agreement among all of the 10 observers. Cohen k test is an index of agreement of each of the 9 observers compared with the
reference observer (observer no. 10). Grade 2, medium size; grade 3, large size varices; red, red marks. CI confidence interval.
0.400.61) and good with Fleiss test (k 0.46, 95% CI 0.4359 Although a small number of studies in children have
0.4943) (Fig. 7). suggested that noninvasive tests may identify subjects at risk of
If small size with red marks, and medium- and large-sized bleeding with sufficient accuracy (5,15,16), endoscopy remains the
varices were adopted in this cohort of children to indicate primary reference test for the diagnosis of varices (4). A survey reported that
prophylaxis, class A would lead to prophylactic treatment in 60% the majority of paediatric gastroenterologists consider endoscopy as
(10%) patients, whereas class B would indicate prophylaxis only the preferred screening modality for oesophageal varices and
in 46% (9%) patients (P 0.005, Student t test). evidence of PH in children with liver disease (17).
The major reason for lack of good quality evidence on
management of PH in children is because many single-centre
DISCUSSION studies have had too few subjects thereby lacking in statistical
Multicentre prospective trials in children with PH and oeso- power. The only way to evaluate strategies to prevent bleeding is to
phageal varices are dependent on the accuracy and reproducibility carry out multicentre studies among centres following children with
of the endoscopic findings (7), yet scoring systems for the diagnosis chronic liver disease and performing diagnostic and therapeutic
and grading of varices have not been validated or standardized in endoscopy. Grading varices is essential in the management of these
children, and little is known about their reproducibility. patients. It follows that validation of this diagnostic procedure is an
Gastrointestinal haemorrhage is an important complication essential preliminary step to any multicentre study (18).
of chronic liver disease or portal vein obstruction in children, and In our study, we selected a group of patients with a clinical
several studies have described effective treatment for oesophageal diagnosis of PH, both cirrhotic and noncirrhotic, consecutively sent
varices in children with cirrhotic and noncirrhotic PH (5,11,12). to endoscopy for evaluation of varices. Endoscopic findings fol-
Nonetheless, there are few published reports providing prevalence lowed a well-balanced distribution making statistics reliable for this
figures derived from routine screening endoscopies for unselected cohort. This is important because a limitation of k is that it is
groups of children with PH. Studies from paediatric hepatology affected by the prevalence of the finding under observation (19).
referral centres suggest that >50% of cirrhotic children have varices We chose to test and compare 2 different classifications to see
(6). The age at the first bleeding episode is related to the which would lead to better agreement among observers. The
underlying aetiology. traditional classification into 3 sizes (class A) performed similarly
In a recent study, 65 children with EHPVO were studied for a to the simplified one (class B), with an overall fair interobserver
median of 8.4 years. A total of 32 (49%) patients presented with agreement in both. This grade of agreement is similar to what
bleeding at a median age 3.8 years, and during the follow-up period has been reported in studies of adult patients with PH (Table 2)
43 of them (66%) had at least 1 bleeding episode (13). A study on (2023).
the long-term outcome of varices in children with EHPVO showed A limitation of this study was represented by the fact that we
that, among 44 children studied from age 12 to a mean age 20 years, used still photographs instead of videos. Although we collected
the probabilities of bleeding by age 18 and 23 years were 60% and pictures taken during insufflation, a 20- to 30-second movie of each
85%, respectively, in patients who had grade II or III oesophageal patient endoscopy would have added power to the observers
varices at age 12, and concluded that children with EHPVO have a evaluation.
risk of bleeding >50% during adolescence (14).
50 70
Agreement in classification A Agreement in classification B
Cohen = 0.34 (range 0.230.45) Cohen = 0.38 (range 0.260.52)
60
40
50
Percentage
Percentage
30
No varices
40
No varices
Small
30 Small
20 Medium
Large
Large
20
10
10
0 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Observers Observe
FIGURE 3. Classification of varices scored by each observer using 3 FIGURE 4. Classification of varices scored by each observer using 2
degrees of size (classification A). The reference observer is number 10. degrees of size (classification B). The reference observer is number 10.
www.jpgn.org 179
Copyright 2015 by ESPGHAN and NASPGHAN. Unauthorized reproduction of this article is prohibited.
DAntiga et al JPGN Volume 61, Number 2, August 2015
80 80
Agreement in classification A grouping medium and large varices Agreement in classification B grouping large and red-marked varices.
Cohen = 0.60 (range 0.490.68) Cohen = 0.46
70 70
60 60
50 50
Percentage
Percentage
40 40
No varices + small No varices + small
Medium + large
30 30
Large + red marks
20 20
10 10
0 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Observers Observers
FIGURE 5. Classification of varices scored by each observer using FIGURE 7. Classification of varices scored by each observer using
classification A and grouping together medium and large versus classification B and grouping together varices of large size plus those
absent and small varices. The reference observer is number 10. with red marks. The reference observer is number 10.
The present guidelines from the Baveno V Consensus Work- indicating primary prophylaxis in adults (medium and large varices,
shop on diagnosis and therapy in PH propose a 3-size classification varices with red marks irrespective of size) were adopted in this
and recommend that patients with small varices without red signs cohort of children, class A would lead to prophylactic treatment in
who have not bled do not require prophylaxis, whereas those with 60% of the patients, an extremely large proportion of cases com-
medium-to-large oesophageal varices and those with red marks, pared with class B that would indicate treatment only in 46%. This
irrespective of size, should benefit from primary prophylaxis (24). It suggests that the observers, when scoring with class B, tend to
should be clarified that in children, there is no evidence to support distribute medium-sized varices unevenly and downgrade them to
primary prophylaxis of oesophageal bleeding for varices of any size the low-risk group. In addition, based on our own data, the 3-grade
or colour. We know that children do not necessarily have the same system showed the worst agreement between small- and medium-
risk of bleeding as adults, and therefore these guidelines cannot be sized varices, so probably here the medium vessels were inappro-
adopted in paediatric patients. priately assigned and with this score the patients would be exposed
Nevertheless, to better understand the differences between to risks of prophylactic treatment inappropriately. For this reason,
class A and class B in this study, we evaluated the agreement and the when it comes to future studies on indication to treat varices in
percentage of patients in our cohort that would have been treated if children, the use of a 3-grade versus a 2-grade score will be
the same indications for prophylaxis in adults were adopted. Group- extremely relevant. From these results, it is clear that further studies
ing medium and large varices against small and no varices in class A to correlate different scoring systems with subsequent risks of
resulted in an excellent agreement, superior to class B. In this bleeding are needed because it is only with such studies that the
perspective, with 3 categories of varices (small, medium, and large), better grading system can be identified.
it would not be relevant if 1 observer classifies varices as medium In 1990, Bendtsen et al (21) carried out an interobserver
and the other as large because the management will be the same for agreement study of varices in adults with cirrhosis. The varices were
both types. Conversely, using class B, if one observer calls the graded on a scale of 0 to 3 according to size. Overall agreement
varices small and the other large, the clinical implication will be among the endoscopists on the presence (grades 1 to 3) or absence
significant because divergent assessment will lead to prophylactic (grade 0) of varices was 70% and was not related to the experience
treatment versus none. Remarkably, if the sizes and the marks
180 www.jpgn.org
Copyright 2015 by ESPGHAN and NASPGHAN. Unauthorized reproduction of this article is prohibited.
JPGN Volume 61, Number 2, August 2015 Endoscopic Classification, Oesophageal Varices
of the endoscopist. The average k value was 0.38 (standard 9. Garcia-Tsao G, Sanyal AJ, Grace ND, et al. Prevention and management
deviation 0.16), showing an overall fair agreement. Discrimi- of gastroesophageal varices and variceal hemorrhage in cirrhosis. Am J
nation between varices with no risk of bleeding (graded 0 to 1) Gastroenterol 2007;102:2086102.
and varices at risk (graded 2 to 3) gave a k value of 0.52. Similarly, 10. Cohen J. A coefficient of agreement for nominal scales. Educ Psychol
Cales et al (20,22,23) performed several studies in adults showing a Meat 1960;20:3746.
good agreement for the presence or the grade of oesophageal varices 11. Duche M, Ducot B, Ackermann O, et al. Experience with endoscopic
and for red signs. These findings are extremely similar to what we management of high-risk gastroesophageal varices, with and without
have demonstrated in the present study focused on children bleeding, in children with biliary atresia. Gastroenterology 2013;145:
(Table 2). Nevertheless, in our opinion, these levels of agreement 801807.
seem suboptimal for such a relevant issue. The variation in grading 12. McKiernan PJ, Beath SV, Davison SM. A prospective study of endo-
of varices shown in our study and previous studies, and therefore the scopic esophageal variceal ligation using a multiband ligator. J Pediatr
Gastroenterol Nutr 2002;34:20711.
potential variation in provision of therapy, is sufficiently concern-
13. Alberti D, Colusso M, Cheli M, et al. Results of a stepwise approach to
ing to require attention to better training of endoscopists to grade
extrahepatic portal vein obstruction in children. J Pediatr Gastroenterol
varices more uniformly. Nutr 2013;57:61926.
In conclusion, in this multicentre study, we found that pae-
14. Lykavieris P, Gauthier F, Hadchouel P, et al. Risk of gastrointestinal
diatric endoscopists working in tertiary care hospitals in Europe are in bleeding during adolescence and early adulthood in children with portal
agreement on the diagnosis and grading of oesophageal varices in vein obstruction. J Pediatr 2000;136:8058.
children. Nevertheless, variation in scoring and therefore in hypothe- 15. Gana JC, Turner D, Mieli-Vergani G, et al. A clinical prediction rule and
tical provision of therapy was wide. The results of this study support platelet count predict esophageal varices in children. Gastroenterology
the feasibility and reliability of multicentre prospective trials in 2011;141:200916.
children with PH, although a better training of paediatric endoscopists 16. Chang HK, Park YJ, Koh H, et al. Hepatic fibrosis scan for liver stiffness
in scoring oesophageal varices is advocated. score measurement: a useful preendoscopic screening test for the
detection of varices in postoperative patients with biliary atresia.
REFERENCES J Pediatr Gastroenterol Nutr 2009;49:3238.
1. DAntiga L. Medical management of esophageal varices and portal 17. Gana JC, Valentino PL, Morinville V, et al. Variation in care for children
hypertension in children. Semin Pediatr Surg 2012;21:2118. with esophageal varices: a study of physicians, patients, and families
2. van Heurn LW, Saing H, Tam PK. Portoenterostomy for biliary atresia: approaches and attitudes. J Pediatr Gastroenterol Nutr 2011;52:7515.
long-term survival and prognosis after esophageal variceal bleeding. 18. Sierra F, Cardenas A. Evidence-based medicine (EBM) in practice:
J Pediatr Surg 2004;39:69. agreement between observers rating esophageal varices: how to cope
3. Shneider BL, Bosch J, de Franchis R, et al. Portal hypertension in with chance? Am J Gastroenterol 2007;102:23636.
children: expert pediatric opinion on the report of the Baveno V 19. Viera AJ, Garrett JM. Understanding interobserver agreement: the
Consensus Workshop on Methodology of Diagnosis and Therapy in kappa statistic. Fam Med 2005;37:3603.
Portal Hypertension. Pediatr Transplant 2012;16:42637. 20. Cales P, Buscail L, Bretagne JF, et al. Interobserver and intercenter
4. Ling SC, Shneider BL. Portal Hypertension in Children: Current Prac- agreement of gastro-esophageal endoscopic signs in cirrhosis. Results of a
tice and the Need for Evidence Oxford: Wiley-Blackwell; 2011: prospective multicenter study. Gastroenterol Clin Biol 1989;13:96773.
189-196. 21. Bendtsen F, Skovgaard LT, Sorensen TI, et al. Agreement among
5. Duche M, Ducot B, Tournay E, et al. Prognostic value of endoscopy in multiple observers on endoscopic diagnosis of esophageal varices
children with biliary atresia at risk for early development of varices and before bleeding. Hepatology 1990;11:3417.
bleeding. Gastroenterology 2010;139:195260.
6. Shneider BL. Approaches to the management of pediatric portal hy- 22. Cales P, Zabotto B, Meskens C, et al. Gastroesophageal endoscopic
pertension: results of an informal survey. In: Groszmann RJ, Bosch J, features in cirrhosis. Observer variability, interassociations, and rela-
eds. Portal Hypertension in the 21st Century. Montreal, Canada: tionship to hepatic dysfunction. Gastroenterology 1990;98:15662.
Springer Netherlands; 2004. 23. Cales P, Burtin P, Person B, et al. Interobserver agreement on endo-
7. Ling SC, Walters T, McKiernan PJ, et al. Primary prophylaxis of scopic diagnosis of diffuse gastric mucosal features in patients with
variceal hemorrhage in children with portal hypertension: a framework cirrhosis. Eur J Gastroenterol Hepatol 1993;5:2638.
for future research. J Pediatr Gastroenterol Nutr 2011;52:25461. 24. de Franchis R. Revising consensus in portal hypertension: report of the
8. Thuluvath PJ, Krishnan A. Primary prophylaxis of variceal bleeding. Baveno V Consensus Workshop on methodology of diagnosis and
Gastrointest Endosc 2003;58:55867. therapy in portal hypertension. J Hepatol 2010;53:7628.
www.jpgn.org 181
Copyright 2015 by ESPGHAN and NASPGHAN. Unauthorized reproduction of this article is prohibited.