Drugs During Pregnancy: Methodological Aspects 1st ed.

10. Lumping or Splitting?

Bengt Källén1

(1)

Tornblad Institute, Lund University, Lund, Sweden

10.1 Lumping or Splitting of Drug Exposures

This can be done on different levels. An example is shown in Table 10.1

Table 10.1

Association between maternal use of antidepressants and infant cardiovascular defects

Drug group

Number of exposures

Number of cardiovascular defects

OR

95 % CI

Any antidepressant

23,658

208

1.04

0.91–1.20

TCA

2139

36

1.29

0.93–1.79

Clomipramine

1399

28

1.49

1.03–2.15

SSRI

19,181

151

0.94

0.80–1.11

Paroxetine

1687

24

1.67

1.12–2.50

From Källén et al. (2013)

Bold text marks statistical significance

TCA tricyclic antidepressant, SSRI selective serotonin reuptake inhibitor

The large group of any antidepressant or the main subgroups, TCA and SSRI, thus do not show any significant risk increase – but the specific antidepressants clomipramine and paroxetine do. If this effect is true, it is hidden in the analysis of larger groups as it is based on only 6 and 7 %, respectively, of the total material. It may, however, be false because when we divide the main group into a number of subgroups, a risk for a “multiple testing” effect exists. The paroxetine association has been seen in some but not all studies performed but is supported by a recent meta-analysis (Bérard et al. 2016).

Sometimes one can lump drugs of quite different nature but with some effect or side effect in common. Thus, for instance, some studies concentrated on folic acid antagonists (Hernández-Diaz et al. 20002001; Matok et al. 2009), others on nitrosatable drugs (Olshan et al. 1989; Gardner et al. 1998). Other similar groupings can be made. A number of drugs with quite different pharmacological effects may cause QT interval prolongation, a mechanism which in experimental systems has been suggested to be teratogenic (Danielsson et al. 2013). A problem in such studies is the inclusion of drugs with a known teratogenic effect, e.g., many anticonvulsants. If they make up a large part of the studied group, they may result in a significance which may be unrelated to the characteristic after which the studied group was selected.

10.2 Lumping or Splitting of Outcomes

Typically this problem arises when congenital malformations are studied. The concept of congenital malformations is so wide and contains conditions with different etiology and embryology. A teratogenic drug could affect a central component in embryonic development like cell division, resulting in widespread malformations, or has a more specific effect like an antiandrogenic effect which could affect male genital development and cause hypospadias.

It is reasonable to study all congenital malformations together. This is actually what the pregnant woman is interested in: does the use of the specific drug she has taken increase the risk for a congenital malformation? She might restrict this to major or serious malformations – she may not be concerned about minor or easily corrected anomalies – but it is still a very heterogeneous concept. To get an effect on the total risk of significance for an individual patient, the drug must affect a common malformation or a group of malformations, e.g., congenital heart defects which make up a substantial part of all major malformations. If the drug selectively doubles the risk for spina bifida, the absolute risk is still very low and an increase in total malformation risk is not detectable.

There is thus a need to group malformations into more homogenous groups in order to identify effects on specific embryonic processes. This has very little to do with what organ the malformation belongs to, the subdivision principle used by the ICD code. This was discussed earlier in this book (Chap. 6). The “lumping” should be made with some consideration of the pathogenetic pathways of different malformations. An example is, for instance, to group neural tube defects (anencephaly, spina bifida, encephalocele) – even though some data actually indicate that anencephaly, encephalocele, and upper spina bifida represent one entity and lower spina bifida another. The former are defects of the closure of the neural tube, the latter origin in the most caudal part of the neural rudiment which develops from the so-called caudal eminence.

Another example is that both esophageal atresia and anal atresia develop by a similar process at about the same time and occur together more often than what chance can explain. Some cases of small gut atresia are probably also related even though other cases may have a quite different pathogenesis.

In order to identify possible specific patterns of malformations among infants exposed to a drug, it is a good idea to list the observed malformations – only in large studies of common drugs will the numbers of malformed infants be too large to permit a detailed listing. If such a listing is made and one notices that a number of cases seem to be identical or resemble each other from a pathogenetic point of view, the statistical evaluation of the observed “cluster” is uncertain. If there was no prior hypothesis to explain the “cluster,” it is not reasonable to test the observed number of cases in the cluster against the expected number. But the cluster should be presented so other investigators can check its validity in their material.

My conclusion is that one should always present the risk of any (at least major) malformation and then present the various malformations by a reasonable grouping, avoiding the “organ system” division. If the material is not too large, it is advisable to present a table with individual malformations (or combinations of malformations), followed with the number of such cases.

An example will be given on limited data after early pregnancy exposure to methimazole (Table 10.2). Only 151 infants were exposed, but 15 (10 %) of them had a congenital malformation diagnosis of variable significance. The risk ratio for any congenital malformation was 2.09 (95 % CI 1.17–3.45), for a relatively severe malformation 2.79 (95 % CI 1.52–4.68).

Table 10.2

Data on malformations observed among 151 infants exposed in utero for methimazole

Malformation

Number

Note

Lacrimal duct stenosis

1

Minor

Ventricular septum defect

1

Major

Single umbilical artery

1

Minor

Unspecified cardiac defect

1

Major

Choanal atresiaa

1

Major

Choanal atresiaa + hypospadias

1

Major

Esophageal atresiaa + ventricular septum

defect + aorta malformation

1

Major

Ileum atresia

1

Major

Meckel diverticlea

1

Minor

Omphalocelea

1

Major

Gastroschisis

1

Major

Hydronephrosis

1

Major

Polydactyly

3

Major

Five of the listed malformations (marked witha) have been mentioned as typical for methimazole embryopathy in the literature (Clementi et al. 1999)

aMalformation regarded as typical for methimazole exposure

References

Bérard A, Iessa N, Chaabane S, Musanda FT, Boukhris T, Zhao J-P (2016) The risk of major cardiac malformations associated with paroxetine during the first trimester. Br J Clin Pharmacol 81(4):589–604. doi:10.1111/bcp.12849CrossRefPubMed

Clementi M, Di Gianantonio E, Pelo E, Mammi I, Basile RT, Tenconi R (1999) Methimazole embryopathy: delineation of the phenotype. Am J Med Genet 83:43–46CrossRefPubMed

Danielsson C, Brask J, Sköld A-C, Genead R, Andersson A, Andersson U et al (2013) Exploration of human, rat, and rabbit embryonic cardiomyocytes suggests K-channel block as a common teratogenic mechanism. Cardiovasc Res 97:23–32CrossRefPubMed

Gardner JS, Guyard-Boileau B, Alderman BW, Fernbach SK, Greene C, Mangione EJ (1998) Maternal exposure to prescription and non-prescription pharmaceuticals or drugs of abuse and risk of craniosynostosis. Int J Epidemiol 27:64–67CrossRefPubMed

Hernández-Dias S, Werler MM, Walker AM, Mitchell AA (2000) Folic acid antagonists during pregnancy and the risk of birth defects. N Engl J Med 343:1608–1614CrossRef

Hernández-Diaz S, Werler MM, Walker AM, Mitchell AA (2001) Neural tube defects in relation to use of folic acid antagonists during pregnancy. Am J Epidemiol 153:961–968CrossRefPubMed

Källén B, Borg N, Reis M (2013) The use of central nervous system active drugs during pregnancy. Pharmacoceuticals 6:1221–1286CrossRef

Matok I, Gorodischer R, Koren G, Landau D, Wiznitzer A, Levy A (2009) Exposure to folic acid antagonists during the first trimester of pregnancy and the risk of major malformations. Brit J Clin Pharmacol 68:956–962CrossRef

Olshan AF, Faustman EM (1989) Nitrosatable drug exposure during pregnancy and adverse pregnancy outcome. Int J Epidemiol 18:891–899CrossRefPubMed