Mr Arbyn

2. Marc Arbyn
Coordinator of the Unit of Cancer Epidemiology

3. Rue Juliette Wytsmanstraat 14 | 1050 Brussels | Belgium
T +32 2 642 50 73 | F +32 2 642 54 10 | email: amidu.raifu@iph.fgov.be | http://www.iph.fgov.be
M Arbyn1
(1) Unit of Cancer Epidemiology, Belgian Cancer Centre
Scientific Institute of Public Health, Brussels, Belgium;
Standards for Colpscopy:
Evidence for the margin status as
indicator for good clinical practice

4. 4

5. History of meta-analysis: margin
status & treatment failure
 Surgical treatment of cervical precancer
(CIN2/3): acceptance for positive margins ≤
20% (= EFC quality indicator: Moss et al,
EJOGRB 2014)
 Meta-analyses
 Ghaem-Maghami, Lancet Onol 2007: significant ↑
risk of residual/recurrent CIN2+
 Arbyn, BMJ 2008; Kyrgiou, BMJ 2016: risk of
preterm delivery (PTD), associated with surgical
treatment of CIN, ↑ with ↑ depth
5

6. History of meta-analysis: margin
status & treatment failure
 Prior meta-analyses
 Arbyn, Vaccine 2012: post-treatment HPV
testing, accurate predictor of residual CIN2+
 Current
 % + margins, % failure & association with margin
 Accuracy of margin status
 Accuracy of post-treatment HPV
 Relative accuracy: margins vs HPV vs both
 Margin status a good quality indicator?
6

7. Meta-analysis: margins & treatment failure
(inclusion criteria)
7
 Restrictions:
 Histological verification of CIN2+ at treatment
 Histological verification of failing cases: CIN2+
 HPV testing between 3-9 months post-treatment
 Follow-up of at least 18 month
 Covariates:
 Treatment procedure
 Grade treated CIN
 Location of involved margins: ecto-cervical,
endo-cervical or both
 Year of publication

8. Meta-analysis: margins & treatment failure
8
Total 97 studies
Risk of treatment failure
associated with margin
status
93 studies
Accuracy studies margin status
25 studies
• Only margin status: 7 studies
• Also post-treatment HPV: 18 studies
• Joint accuracy of margin status &
post-treatment HPV: 5 studies
1. Meta-analysis Ghaem-Maghani, 2007 [residual CIN2+ ~ margins]: 65 studies
2. Meta-analyses Arbyn, Vaccine2006/12 & update [post-treatment HPV]: 15 studies
3. New searches (2006-16): 16 studies
Excluded:
 Double reporting: 1
 Case-control studies: 3

9. Subtotal (I2 = 95.4%, p< 0.001)
Andersen, 1994
Kong, 2014
Mazouni, 2005
Bandieramonte, 1998
Maluf, 2004
Ueda, 2006
Fambrini, 2008
Vedel, 1993
Vergote, 1992
Mohamed-Noor, 1997
Reich, 2002
Skjeldestad, 1997
Orbo, 2004
Grundsell, 1983
LC
Subtotal (I2 = 98.3%, p< 0.001)
Izumi, 2000
Stamatopoulos, 2001
Lopes, 1993
Moore, 1992
Ahlgren, 1975
Bertelsen, 1999
Murta, 1999
Milojkovic, 2002
White, 1993
Murta, 2004
Demopoulos, 1991
Riethmuller, 2008
Paterson-Brown, 1992
Larsson, 1981
Orbo, 2004
Bjerre, 1976
CKC
Study
16.9 (13.8, 20.6)
13.3 (11.0, 16.1)
15.2 (12.2, 18.8)
2.8 (1.1, 6.9)
46.6 (34.3, 59.2)
12.3 (10.9, 13.8)
15.4 (8.0, 27.5)
19.2 (15.6, 23.5)
26.5 (18.8, 36.0)
18.2 (15.5, 21.2)
8.1 (7.4, 8.9)
23.4 (20.9, 26.0)
27.3 (19.1, 37.4)
7.1 (4.7, 10.7)
17.8 (12.9, 23.2)
20.8 (13.1, 31.6)
12.4 (8.1, 18.6)
42.7 (37.3, 48.3)
19.8 (15.7, 24.7)
13.5 (11.0, 16.6)
35.1 (27.5, 43.6)
4.1 (3.0, 5.5)
18.8 (13.3, 25.8)
41.4 (33.7, 49.5)
16.7 (12.6, 21.9)
24.6 (20.6, 29.1)
40.3 (34.6, 46.2)
6.1 (4.5, 8.0)
36.1 (33.6, 38.7)
ES (95% CI)
20.2 (14.3, 26.7)
21.4 (14.8, 29.9)
11.9 (9.0, 15.5)
0 20 40 60 80 100
% positive margins
Overall (I2 = 95.8%, p< 0.001)
Dobbs, 2000
Ang, 2011
Alonso, 2006
Ribaldone, 2010
Kang, 2010
Wu, 2016
Mints, 2006
Gardeil, 1997
Herfs, 2015
Bodner, 2002
Robinson, 1998
Aerssens, 2009
Fuste, 2009
Prato, 2008
Verguts, 2006
Goff, 1994
Gonzalez, 2001
Bae, 2007
Skinner, 2004
Hulman, 1998
Johnson, 2003
Felix, 1994
Nagai, 2004
de Cabezon, 1998
Jain, 2001
Kucera, 2001
Flannelly, 2001
Zaitoun, 2000
Brismar, 2009
Lu, 2006
Shafi, 1993
Livasy, 1999
Kang, 2016
Houfflin, 2003
Baldauf, 1998
Ryu, 2012
Spitzer, 1993
Gosvig, 2015
Bar-Am, 2000
Torne, 2013
Hanau, 1997
Hallam, 1993
Study
25.9 (22.3, 29.6)
23.4 (19.1, 28.3)
47.3 (44.8, 49.8)
32.8 (26.7, 39.6)
10.3 (5.3, 19.0)
16.5 (13.9, 19.5)
13.2 (11.1, 15.7)
8.8 (5.2, 14.4)
47.5 (40.8, 54.4)
26.0 (19.2, 34.1)
16.2 (7.7, 31.1)
46.7 (38.1, 55.5)
16.4 (10.9, 24.0)
29.6 (22.0, 38.5)
19.4 (12.0, 30.0)
16.7 (10.7, 25.1)
31.7 (25.0, 39.2)
21.1 (14.6, 29.4)
39.5 (35.1, 44.0)
31.4 (27.5, 35.6)
32.0 (28.6, 35.6)
28.1 (18.1, 40.8)
5.6 (2.9, 10.7)
27.1 (18.1, 38.5)
59.5 (48.5, 69.6)
12.7 (8.2, 19.1)
30.1 (28.3, 31.9)
41.2 (38.3, 44.2)
28.2 (19.8, 38.6)
30.1 (26.0, 34.5)
20.9 (15.2, 28.0)
53.0 (46.1, 59.8)
10.7 (7.2, 15.6)
36.1 (29.8, 42.9)
16.5 (12.5, 21.4)
26.2 (20.4, 33.0)
17.4 (12.5, 23.8)
23.1 (19.8, 26.8)
9.5 (5.6, 15.6)
22.7 (16.4, 30.6)
25.3 (17.3, 35.3)
28.4 (25.6, 31.5)
42.9 (33.8, 52.4)
30.1
17.4 (12.5, 23.8)
ES (95% CI)
0 20 40 60 80 100
% positive margins
Proportion of involved margins (%)
CKC: 20.2%
LC: 17.8%
LLETZ: 25.9%

10. Trend in % positive margins (by treatment procedure)

11. Heterogeneity between groups: p < 0.001
Overall (I2 = 92.3%, p<0.0001)
Brismar, 2009
Subtotal (I2 = 90.1%, p<0.0001)
Herfs, 2015
Nagai, 2004
Study
Alonso, 2006
Zhao, 2014
Guerra, 1996
Reich, 2002
Gallwas, 2010
LLETZ
Torne, 2013
Ang, 2011
LC
Subtotal (I2 = 35.7%, p=0.17)
Chao, 2004
Verguts, 2006
Kang, 2010
Trope, 2011
Kang, 2016
Aerssens, 2009
Skinner, 2004
Orbo, 2004
Gosvig, 2015
Subtotal (I2 = 99.7%, p<0.0001)
Fambrini, 2008
Ryu, 2012
Bandieramonte, 1998
Murta, 1999
Wu, 2016
Subtotal (I2 = 99.7%, p<0.0001)
CKC
Mixed
6.6 (4.9, 8.4)
5.9 (2.5, 13.0)
6.7 (4.6, 9.3)
3.1 (1.2, 7.6)
ES (95% CI)
11.9 (8.2, 17.2)
6.8 (5.4, 8.5)
2.1 (1.7, 2.5)
10.3 (5.8, 17.5)
9.1 (5.3, 15.2)
3.7 (2.8, 4.7)
6.8 (5.7, 8.1)
6.0 (4.5, 7.9)
9.7 (4.8, 18.7)
5.5 (4.0, 7.5)
6.4 (4.3, 9.5)
12.6 (8.8, 17.9)
9.8 (5.7, 16.4)
15.4 (12.3, 18.9)
9.2 (6.8, 12.1)
2.9 (1.8, 4.7)
2.1 (0.4, 4.9)
5.8 (2.0, 15.6)
6.6 (3.8, 11.1)
2.2 (1.4, 3.4)
2.2 (1.8, 2.6)
4.9 (2.4, 9.8)
5.2 (3.2, 8.1)
1.4 (0.4, 4.9)
18.3 (12.6, 25.8)
0 10 20 30 40 50
Residual/recurrent CIN2+ (%)
Occurrence of residual/recurrent CIN2+ (=treatment failure)
% Failure: 6.6%
(95% CI: 4.9-8.4)

12. 0.5 1 2 10 100
Risk CIN2+ ~ margin status
Heterogeneity between groups: p = 0.002
Overall (I2 = 89.3%, p<0.001);
Study
Orbo, 2004
Subtotal (I^2 = 0.0%, p = 0.5)
Aerssens, 2009
LC
Brismar, 2009
Zhao, 2014
Subtotal (I^2 = 73.6%, p = 0.0)
LLETZ
Herfs, 2015
Mixed
Ryu, 2012
Skinner, 2004
Alonso, 2006
Subtotal (I^2 = 91.0%, p = 0.0)
Subtotal (I^2 = 0.0%, p = 0.5)
Ahlgren, 1975
Torne, 2013
Trope, 2011
Nagai, 2004
Reich, 2002
Kang, 2010
Fambrini, 2008
Orbo, 2004
Bandieramonte, 1998
Chao, 2004
Ang, 2011
CKC
Gallwas, 2010
Dobbs, 2000
Larsson, 1981
Bjerre, 1976
Verguts, 2006
Orbo, 2004
Murta, 1999
17.5 (12.8, 22.7)
ES (95% CI)
11.8 (6.1, 21.5)
12.6 (10.0, 15.4)
15.0 (5.2, 36.0)
8.3 (2.3, 25.8)
14.7 (10.7, 19.9)
25.6 (19.6, 32.2)
8.8 (3.0, 23.0)
20.8 (11.7, 34.3)
28.3 (22.3, 35.3)
27.3 (18.0, 39.0)
15.9 (8.3, 25.2)
14.1 (3.0, 29.5)
30.0 (19.9, 42.5)
33.3 (19.2, 51.2)
19.7 (11.6, 31.3)
12.5 (2.2, 47.1)
21.5 (17.7, 25.9)
17.1 (11.2, 25.2)
25.0 (7.1, 59.1)
16.7 (6.7, 35.9)
0.0 (0.0, 49.0)
11.5 (8.2, 15.7)
3.9 (2.8, 5.6)
10.0 (1.8, 40.4)
12.0 (6.4, 21.3)
40.9 (27.7, 55.6)
24.6 (21.0, 28.6)
14.3 (4.0, 39.9)
9.1 (3.6, 21.2)
37.0 (24.5, 51.4)
ES (95% CI)
11.8 (6.1, 21.5)
12.6 (10.0, 15.4)
15.0 (5.2, 36.0)
8.3 (2.3, 25.8)
14.7 (10.7, 19.9)
25.6 (19.6, 32.2)
8.8 (3.0, 23.0)
20.8 (11.7, 34.3)
28.3 (22.3, 35.3)
27.3 (18.0, 39.0)
15.9 (8.3, 25.2)
14.1 (3.0, 29.5)
30.0 (19.9, 42.5)
33.3 (19.2, 51.2)
19.7 (11.6, 31.3)
12.5 (2.2, 47.1)
21.5 (17.7, 25.9)
17.1 (11.2, 25.2)
25.0 (7.1, 59.1)
16.7 (6.7, 35.9)
0.0 (0.0, 49.0)
11.5 (8.2, 15.7)
3.9 (2.8, 5.6)
10.0 (1.8, 40.4)
12.0 (6.4, 21.3)
40.9 (27.7, 55.6)
24.6 (21.0, 28.6)
14.3 (4.0, 39.9)
9.1 (3.6, 21.2)
37.0 (24.5, 51.4)
0 20 40 60 80 100
PPV (%)
Margin+
Heterogeneity between groups: p = 0.319
Overall (I2 = 92.3%, p<0.001)
Brismar, 2009
Bjerre, 1976
Alonso, 2006
Kang, 2010
Verguts, 2006
Subtotal (I^2 = 97.1%, p = 0.0)
Gallwas, 2010
Subtotal (I^2 = 62.7%, p = 0.0)
Subtotal (I^2 = 0.0%, p = 0.7)
LC
Mixed
Torne, 2013
Ryu, 2012
Zhao, 2014
Subtotal (I^2 = 80.5%, p = 0.0)
Aerssens, 2009
Ang, 2011
Nagai, 2004
Ahlgren, 1975
Trope, 2011
Dobbs, 2000
Orbo, 2004
LLETZ
Orbo, 2004
Murta, 1999
Fambrini, 2008
Chao, 2004
Skinner, 2004
Bandieramonte, 1998
Orbo, 2004
Herfs, 2015
Larsson, 1981
Reich, 2002
CKC
Study
3.8 (2.4, 5.4)
4.9 (1.7, 13.5)
3.0 (2.1, 4.4)
4.4 (2.1, 9.4)
3.2 (2.0, 5.0)
8.6 (3.7, 18.6)
3.5 (0.9, 7.7)
10.3 (5.7, 17.9)
3.5 (2.3, 4.9)
1.8 (0.3, 4.1)
2.0 (0.5, 6.9)
1.5 (0.4, 5.2)
4.4 (3.1, 6.1)
5.3 (3.1, 7.9)
8.8 (4.7, 15.9)
3.4 (2.4, 4.9)
4.4 (2.1, 9.4)
0.4 (0.1, 2.3)
3.5 (1.9, 6.4)
0.8 (0.2, 2.9)
3.1 (0.9, 10.7)
8.7 (6.3, 11.9)
8.2 (4.0, 16.0)
2.3 (0.4, 11.8)
2.9 (1.7, 4.8)
6.9 (4.5, 10.5)
1.4 (0.4, 5.1)
9.8 (7.0, 13.5)
1.0 (0.2, 5.6)
5.7 (4.2, 7.7)
0.4 (0.2, 0.6)
ES (95% CI)
4.9 (1.7, 13.5)
3.0 (2.1, 4.4)
4.4 (2.1, 9.4)
3.2 (2.0, 5.0)
8.6 (3.7, 18.6)
3.5 (0.9, 7.7)
10.3 (5.7, 17.9)
3.5 (2.3, 4.9)
1.8 (0.3, 4.1)
2.0 (0.5, 6.9)
1.5 (0.4, 5.2)
4.4 (3.1, 6.1)
5.3 (3.1, 7.9)
8.8 (4.7, 15.9)
3.4 (2.4, 4.9)
4.4 (2.1, 9.4)
0.4 (0.1, 2.3)
3.5 (1.9, 6.4)
0.8 (0.2, 2.9)
3.1 (0.9, 10.7)
8.7 (6.3, 11.9)
8.2 (4.0, 16.0)
2.3 (0.4, 11.8)
2.9 (1.7, 4.8)
6.9 (4.5, 10.5)
1.4 (0.4, 5.1)
9.8 (7.0, 13.5)
1.0 (0.2, 5.6)
5.7 (4.2, 7.7)
0.4 (0.2, 0.6)
ES (95% CI)
0 20 40 60 80 100
cNPV (%)
Margin-
.
.
.
.
Overall (I2 = 85.9%, p=0.000)
Bandieramonte, 1998
Mixed
Subtotal (I-squared = 70.2%, p = 0.000)
CKC
Orbo, 2004
Alonso, 2006
Subtotal (I-squared = 56.8%, p = 0.055)
Verguts, 2006
Kang, 2010
Torne, 2013
ID
Ryu, 2012
Study
Ang, 2011
Trope, 2011
Nagai, 2004
LC
Murta, 1999
Gallwas, 2010
Aerssens, 2009
Larsson, 1981
Dobbs, 2000
Subtotal (I-squared = 0.0%, p = 0.873)
Herfs, 2015
Orbo, 2004
LLETZ
Zhao, 2014
Chao, 2004
Skinner, 2004
Subtotal (I-squared = 93.4%, p = 0.000)
Brismar, 2009
Orbo, 2004
Bjerre, 1976
Fambrini, 2008
Ahlgren, 1975
Reich, 2002
4.9 (3.1, 7.6)
5.6 (0.3, 102.7)
4.2 (2.4, 7.2)
0.9 (0.3, 2.5)
6.1 (2.6, 14.7)
3.0 (1.8, 4.9)
1.7 (0.4, 7.7)
5.3 (2.9, 9.8)
17.0 (3.9, 73.4)
ratio (95% CI)
14.1 (3.2, 61.9)
1.2 (0.7, 1.9)
5.6 (2.5, 12.3)
2.8 (0.4, 20.6)
4.5 (2.0, 10.0)
1.0 (0.1, 6.8)
1.7 (0.5, 5.7)
7.2 (4.5, 11.4)
14.8 (3.3, 66.8)
6.6 (2.0, 22.0)
8.6 (0.9, 79.5)
1.4 (0.7, 2.8)
3.4 (2.1, 5.3)
4.0 (2.2, 7.3)
4.1 (2.5, 6.7)
9.2 (3.3, 25.7)
1.7 (0.3, 9.5)
5.3 (1.0, 27.3)
8.1 (5.4, 12.2)
11.0 (1.1, 107.5)
72.9 (9.9, 535.3)
56.0 (33.6, 93.3)
5.6 (0.3, 102.7)
4.2 (2.4, 7.2)
0.9 (0.3, 2.5)
6.1 (2.6, 14.7)
3.0 (1.8, 4.9)
1.7 (0.4, 7.7)
5.3 (2.9, 9.8)
17.0 (3.9, 73.4)
ratio (95% CI)
14.1 (3.2, 61.9)
1.2 (0.7, 1.9)
5.6 (2.5, 12.3)
2.8 (0.4, 20.6)
4.5 (2.0, 10.0)
1.0 (0.1, 6.8)
1.7 (0.5, 5.7)
7.2 (4.5, 11.4)
14.8 (3.3, 66.8)
6.6 (2.0, 22.0)
8.6 (0.9, 79.5)
1.4 (0.7, 2.8)
3.4 (2.1, 5.3)
4.0 (2.2, 7.3)
4.1 (2.5, 6.7)
9.2 (3.3, 25.7)
1.7 (0.3, 9.5)
5.3 (1.0, 27.3)
8.1 (5.4, 12.2)
11.0 (1.1, 107.5)
72.9 (9.9, 535.3)
56.0 (33.6, 93.3)
RR
Relative risk
17.5% 3.8% 4.9

13. Location of margin involvement
13
 Positivity rate:
 Unspecified: 23.1%
 Ectocervical: 10.4%
 Endocervical: 11.0%
 Endo- & ecto-cervical: 2.9%
 Occurrence of residual/recurrent CIN2+:
 Ectocervical: 7.2% (95% CI 0.0-23.6%)
 Endocervical: 16.3% (95% CI 5.9-29.9%)
 Endo- & ecto-cervical: 18.9% (95% CI 0.0-62.9%)

14. Accuracy to predict residual or
recurrent CIN2+
14
Margin status vs post-treatment HPV testing

15. Quality of diagnostic test accuracy studies (QUADAS tool)
Marginal & HPV post-treatment status to predict treatment outcome
(18 studies)
P1 P2 P3 T1 T2 R1 R2 R3 F1 F2 F3 F4 F5 F6
Chua,199726
Y Y Y Y Y Y U Y Y Y U Y Y Y
Chao,2004
61 Y Y Y Y U Y N Y Y Y N Y U U
Nagai, 2004 65
Y Y Y Y Y Y N Y Y N N Y Y U
Alonso, 200670
Y Y Y Y Y Y U Y Y Y N Y Y U
Verguts,200676
Y Y Y Y Y Y N Y Y Y Y Y Y Y
Fambrini,200878
Y Y Y Y Y Y Y Y Y Y Y Y Y Y
Aerssens,200981
Y Y Y Y Y Y N Y Y Y Y Y Y Y
Brismar, 200982
Y Y N Y Y ? Y Y N U U Y Y Y
Kang, 201087
Y Y Y Y Y Y N Y Y N N Y U U
Trope,201192
Y Y U Y Y Y N Y Y Y Y Y Y
Ryu,201294
Y Y Y Y Y Y Y Y Y Y Y Y Y Y
Torne,201396
Y Y Y Y Y Y N Y Y Y ? Y Y Y
Kong, 201497
Y Y Y Y Y Y U Y N N N Y U U
Zhao, 201498
Y Y Y Y Y Y Y Y Y Y Y Y U U
Gosvig, 201599
Y Y Y Y Y Y U Y Y Y U Y Y Y
Herfs, 2015100
Y Y Y Y U Y Y Y Y Y Y Y Y Y
Kang, 2016101
Y Y Y Y U Y N Y Y Y Y U U U
Wu, 2016102
Y Y Y Y U Y U Y Y U Y N U N
% yes 100 100 88 100 78 94 28 100 89 72 50 88 67 55
% unclear 0 0 6 0 22 6 28 0 0 11 22 6 33 39
% no 0 0 6 0 0 0 44 0 11 17 28 6 0 6
Author, Year
Risk of Bias
Patient Selection Compared tests Reference Test Flow & Timing

16. 020.406080100
Sensitivity(%)
020406080100
Specificity (%)
hrHPV
margin status
Accuracy of margin status and post-treatment
HPV testing to predict treatment failure
Sensi Speci
Margins 55.8% 84.4%
HPV 91.0% 83.8%

17. Accuracy of margin status and post-treatment
HPV testing to predict treatment failure
Absolute Absolute
sensitivity (95% CI) specificity (95% CI)
Margins 55.8%(45.8-65.5%) 84.4%(79.5-88.4%)
hrHPV 91.0%(82.3-95.5%) 83.8%(77.7-88.7%)
margins Relative sensitivity Relative specificity
vs HPV 0.62 (0.53-0.72) 1.01 (0.97-1.06)

18. 0.01
0.05
0.1
0.5
1
5
10
20
50
80
90
95
99
99.5
99.9
Post-testProbability(%)
0.01
0.05
0.1
0.5
1
5
10
20
50
80
90
95
99
99.5
99.9
Pre-testProbability(%)
Pre
(6.6%)
Post(+)
(28.4%)
Post(-)
(0.8%)
0.01
0.05
0.1
0.5
1
5
10
20
50
80
90
95
99
99.5
99.9
Post-testProbability(%)
20
0.01
0.05
0.1
0.5
1
5
10
50
80
90
95
99
99.5
99.9
Pre-testProbability(%)
Pre
(6.6%)
Post(+)
(17.1%)
Post(-)
(3.7%)
Margin status hrHPV
Clinical utility of a test: PPP plot (pretest-posttest probability)

19. 020406080
Positivemargin(%)
CKC LC LLETZ MIX
<20% + margin status

20. Margin status as quality indicator of good
clinical practice
05101520
%ofstudies
0 20 40 60 80
Proportion involved margins (%)

21. 21
Conclusion: prediction of cure/failure
 Women with positive margins show increased
risk of treatment failure
 Post-treatment HPV testing: more accurate
than margin status
 Margin status is not an accurate quality
indicator in colposcopy practice
 Not even one study addresses both outcomes
(obstetrical oncological safety) => object for
new research)
 Update of EU guidelines on management of
screen+ women needed

22. 22
Margin status a good quality indicator
for clinical practice?
 Pertinent: (+++) RR
 Reproducible: (-/+) broad
 Accurate: (+)
 Actionable: (+), may induce obstetrical harm
 Conclusion: poor to moderate evidence

23. 23
Acknowledgements
 European Federation of Colposcopy
 Institut national du Cancer
 European Commission: CoHear(FP7)