EUROCAT Post-marketing
surveillance for congenital
anomalies
Dr. Johannes Michiel Luteijn
Prof. Joan Morris
BINOCAR Scientific Meeting
Swansea, October 7th, 2014
Thanks to:
Lolkje de Jong-van den Berg
Helen Dolk
Ester Garne
Joanne Given
Maria Loane
Drug use in pregnancy

Two studies estimated 57-79% of pregnant women in Northern
Europe are exposed to prescription drugs during pregnancy.1,2

Little is known about the safety of drugs in pregnancy, partly due
to pregnant women not being involved in pre-marketing studies.
1 Bakker, M. K., Jentink, J., Vroom, F., Van Den Berg, P. B., De Walle, H. E., & De Jong-Van Den Berg, L. T. (2006). Drug prescription patterns before, during and after
pregnancy for chronic, occasional and pregnancy-related drugs in the netherlands. BJOG : An International Journal of Obstetrics and Gynaecology, 113(5), 559-568
2 Engeland, A., Bramness, J. G., Daltveit, A. K., Ronning, M., Skurtveit, S., & Furu, K. (2008). Prescription drug use among fathers and mothers before and during
pregnancy. A population-based cohort study of 106,000 pregnancies in norway 2004-2006. British Journal of Clinical Pharmacology, 65(5), 653-660
EUROmediCAT signal detection

Aim:
To develop a method of signal detection to be used as a
routine analysis tool.
Dataset





15 EUROCAT registries in EUROmediCAT
Births from 1995-2011
14,950 registrations of congenital anomaly with at least a
single exposure to a drug (ATC-code 5-digit+)
Folic acid, minerals and vitamins not included as exposures
Exclusions
 Anomalies with 1-4 digit ATC-code exposures only (n=1,288)
 Anomalies with no clear 1st trimester exposure (n=13,377)
 Infants with a chromosomal anomaly
Dataset

Drug exposure
 5- and 7-digit ATC codes (n = 272, n = 505)
 Exposure collected mainly prospective to birth by healthcare
professionals.

Cases
 EUROCAT coding committee identified 59 subgroups of nonchromosomal congenital anomaly
Potential drug-congenital anomaly signals must
have at least 3 exposed cases
3 Teratogenic mechanisms of medical drugs. Van Gelder MM, van Rooij IA, Miller RK, Zielhuis GA, de Jong-van den Berg LTW,
Roeleveld N. Hum Reprod. Update. 2010 Jul-Aug;16(4):378-94
Analysis of drug-CA combinations

Analysis by Fishers’ exact test
Drug A
All other drugs
excluding A
Anomaly
a
b
All other anomalies
c
d

A total of 59 congenital anomalies x 836 drug-codes.

49,324 analyses performed.
 19,529 5-digit ATC-code analyses
 29,795 7-digit ATC-code analyses
Fisher’s exact results for 7-digit ATC –
anomaly combinations
1.00e-10
1.00e-09
1.00e-08
1.00e-07
p
1.00e-06
.00001
P-value .0001
.001
.01
0.05
.05
.1
1
0.01
0.1 0.2
1
5 10
Observed
ORRatio
Observed
Odds
100
Multiple testing issues




P Values
Given there is no association - P = probability observed or more
extreme result would occur
 P = 0.05 means that 5% of Fisher’s exact tests where there is no
association will be “statistically significant”
 Ie 5% of 5,000 tests are likely to be false positives
False Discovery Rate:
Proportion of significant results that are expected to be false
positives
 FDR of 5% means 5% of the “significant” results are likely to be
false positives
 Ie 5% of significant results are likely to be false positives
False Discovery Rate
Density
Distribution of test
statistics if there are no
associations
B
Possible test results (Z-score)
False Discovery Rate
𝐻𝑒𝑖𝑔ℎ𝑡 𝑏𝑙𝑎𝑐𝑘 𝑐𝑢𝑟𝑣𝑒
𝐻𝑒𝑖𝑔ℎ𝑡 𝑟𝑒𝑑 𝑐𝑢𝑟𝑣𝑒
= 𝑩/𝑨
Distribution of test
statistics if there are no
associations
Density
FDR =
Observed distribution
of test statistics
A
B
Possible test results (Z-score)
False Discovery Rate
1.00e-10
1.00e-09
1.00e-08
1.00e-07
p
1.00e-06
.00001
FDR = 5%
.0001
P-value
.001
.01
0.05
.05
.1
0.05
1
0.01
0.1 0.2
1
5 10
Observed
OR
Observed Odds Ratio
100
False Discovery Rate
1.00e-10
May exclude
true
associations
1.00e-09
1.00e-08
1.00e-07
p
1.00e-06
.00001
FDR = 5%
.0001
P-value
.001
.01
0.05
.05
.1
0.05
1
0.01
0.1 0.2
1
5 10
Observed
OR
Observed Odds Ratio
100
False Discovery Rate
1.00e-10
May exclude
true
associations
1.00e-09
1.00e-08
1.00e-07
p
1.00e-06
.00001
FDR = 5%
.0001
P-value
FDR = 50%
.001
.01
0.05
.05
.1
1
0.01
0.1 0.2
1
5 10
Observed
OR
Observed Odds Ratio
100
False Discovery Rate
1.00e-10
May exclude
true
associations
1.00e-09
1.00e-08
1.00e-07
p
1.00e-06
.00001
FDR = 5%
.0001
P-value
FDR = 50%
.001
May include
.05 too many
false
associations
.01
0.05
.1
1
0.01
0.1 0.2
1
5 10
Observed
OR
Observed Odds Ratio
100
Flowchart of analyses
49,324 drug-CA combinations
19,529 5-digit drug combinations
(including aggregate drug groups)
29,795 7-digit drug combinations
Multiple testing procedure
(649 p-value < 0.05)
Multiple testing procedure
(886 p-value < 0.05)
97 potential signals at
(FDR= 0.5 ; p < 0.003)
31 potential signals at
(FDR = 0.5 ; p < 0.0005)
128 potential signals
Preliminary Results – Association tree
Example – Spina bifida and N03AG01 (valproic acid)
Congenital
Anomaly
Neural Tube
Defects
Spina Bifida
Exposure
Neural Tube
Defects
Spina Bifida
Odds Ratio
N03AG
Exposed
Cases
30
4.1 (2.6 - 6.0)
N03AG
29
7.4 (4.7 - 11.2)
N03AG01
29
4.0 (2.6 - 6.0)
N03AG01
28
7.3 (4.6 - 11.0)
Cases not yet verified by registry
Flowchart of analyses
49,324 drug-CA combinations
Exclusions
47,789 FDR > 0.5
128 potential signals
13 Less than 3 exposed cases
115 potential signals
20 5-digit ATC code of 7-digit ATC
associated with the same CA
6 Aggregate CA of a more specific
CA associated with the same
exposure
12 Protective associations
77 Independent signals
Subject to detailed follow-up
Validation of methodology

Spina bifida and valproic acid †
 OR 7.25 (4.61-11.02)
 28 Exposed cases

Congenital heart defects and human insulin †
 OR 2.46 (1.81-3.35)
 102 Exposed cases
† Cases not yet verified by registry
Future work

The WHO-Uppsala Monitoring Centre receives reports of
adverse drug reactions from all over the world.
 66,000 reports of CA-related adverse drug reactions
since 2000
 70% American/Canadian

Valuable independent dataset for testing signals