1 / 38

Challenges in research on drug exposure in pregnancy

Challenges in research on drug exposure in pregnancy. Henrik Toft Sørensen Department of Clinical Epidemiology Aarhus University Hospital Denmark hts@dce.au.dk. Historical Background. Birth defects are part of the human condition observed throughout history.

dewei
Download Presentation

Challenges in research on drug exposure in pregnancy

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Challenges in research on drug exposure in pregnancy Henrik Toft Sørensen Department of Clinical Epidemiology Aarhus University Hospital Denmark hts@dce.au.dk

  2. HistoricalBackground • Birth defects are part of the human condition observed throughout history. • Major birth defects affect approximately 3-4% of live-born infants. • Just over 50 year ago, it was believed that the placenta protected the foetus from noxious agents.

  3. The cause of birth defects Ref : Gale Encyclopedia of Medicine, 2002

  4. Historical Background • This belief was shattered by the recognition in 1941 that maternal rubella infection produced a distinctive pattern of birth defects. • In 1961 the thalidomide disaster demonstrated that drugs could be teratogenic.

  5. Historical Background • 16 December 1961 • McBride’s letter-to-the-editor (15 lines) in the Lancet HTS5

  6. Historical Background • Thalidomide • Thalidomide was a mild hypnotic drug marketed at the end of 1950s • It was used in pregnancy specifically because of its safety profile • Phocomelia – the absence of limbs or parts of limbs • (20- 30%)

  7. More than 40 years after the recognition of thalidomide-associated embryopathy: • More than 70% of all pregnant women used drugs during pregnancy. • Overall, the prevalence of birth defects is stable or decreasing • Fewer than 30-40 drugs have been proved to be teratogenic in humans. • However, little is known about the effect on birth weight, fetal growth and birth outcomes and the consequences for the children

  8. Why Epidemiological Studies? • For most known human teratogens (including thalidomide) results of animal tests vary so much as to seriously limit their predictive value. • Structure or activity of the drug are generally not predictive of teratogenesis. • Thalidomide and glutethimide are structurally closely related, but there is no evidence that the latter is teratogenic. • Pregnant women are excluded from premarketing studies and clinical trials because of fear of teratogenicity.

  9. Epidemiologic issues involved in the study of birth defects are similar to those of other adverse outcomes, but the following considerations are especially important for birth defects: • Sample size • Definition and classification of exposure and outcome • Confounding • Biologic plausibilities

  10. Birth defects cannot be considered as a single homogenous outcome. • Physical birth defects include a wide range of malformations that vary in many ways including: • Gestational timing • Embryological tissue of origin • Mechanism of development

  11. Examples: • Chromosomal abnormalities generally preclude conception • Neural tube defects occur in the earliest week of gestation • Cleft palate occurs toward the end of the first trimester • Microcephaly can develop relatively late in pregnancy

  12. Adult life Birth Conception Morbidity Spontaneous abortions Pre-natal diagnosis of birth defects Stillbirth Birth defects Preterm birth Low birth weight

  13. Teratogens do not uniformly increase the rates of all birth defects but rather increase rates of selected defects. Examples: Thalidomide - limb defects Isotretinoin - ear, central nervous system and cardiac defects Valproic acid - neural tube defects Warfarin - cartilage defects ACE*-inhibitors - renal defects * Angiotensin converting enzyme

  14. The fact that pharmacoepidemiologic studies must consider specific rather than overall rates of birth defects has a dramatic effect on sample size requirements. To detect a doubling risk of a relatively common specific birth defect (1/1000 ~ oral clefts) one would require a sample size of 23 000 exposed pregnancies.

  15. The overall rarity of malformations, particularly of specific defects, requires large cohorts or necessitates resorting to the case-control design in relevant pharmacoepidemiologic studies.

  16. Cohort design The Present The Future Birth defect No birth defect Risk factor present Risk factor absent Birth defect No birth defect Population Sample Example: Rebordosa et al. Use of acetaminophen during pregnancy and risk of adverse pregnancy outcomes. Int J Epidemiol 2009; 38: 706-14.

  17. Case-control design The Past or Present The Present Risk factor present Risk factor absent Birth defect Population with disease (cases) Sample of cases Much larger population without disease (controls) Risk factor present Risk factor absent No birth defect Sample of controls Example: Catonet al. Antihypertensive medication use during pregnancy and the risk of cardiovascular malformations. Hypertension 2009; 54:63-70.

  18. Cohort Studies versus Case-control Studies1 Example of cohort study: Incidence of birth defect among unexposed 0.0005 Relative risk to be detected 2 Beta 0.20 (power 80%) Alpha 0.05 26 824 exposed and 107 296 unexposed needed Unexposed:exposed ratio 4:1

  19. Cohort Studies versus Case-control Studies2 Example of case-control study: Prevalence of exposure in controls 0.05 Relative risk to be detected 2.0 Beta 0.20 (power 80%) Alfa 0.05 516 cases and 516 controls needed Case: control ratio 1:1

  20. Case-control Surveillance • Hungary • The Hungarian Case-Control Surveillance of Congenital Abnormalities • 23 000 cases and 38 000 controls • Study base 1.8m births • Example: Bártfai et al. A population-based case-control teratologic study of promethazine use during pregnancy. Reprod Toxicol 2008: 25:276-85.

  21. Case-control Surveillance 2. USA The Case-control Surveillance, the Slone Epidemiology Unit, Boston > 15 000 babies with birth defects > 1000 cases of neural tube defects > 500 cases of cleft palate > 80 cases of gastroschisis Example: Cator et al. Maternal hypertension, antihypertensive medication use, and the risk of severe hypospadias. Birth Defects Res A Clin Mol Teratol 2008; 82: 34-40.

  22. Problems with Case-control Studies Examples of problems in the Hungarian Case-Control Surveillance of Congenital Abnormalities 1. Most women treated with other drugs – uncontrolled confounders such as e.g. underlying disease 2. Validation study has shown that recall bias is a problem:be careful with odds ratios less than 2. 3. The response rate lower for controls than cases – risk of selection bias Example: Rockenbauer et al. Recall bias in a case-control surveillance system on the use of medicine during pregnancy. Epidemiology 2001;12:461-6.

  23. Conclusion Cohort studies need to be large - therefore cohort studies in pharmacoepidemiology are often based on existing databases or as an alternative Case-control studies

  24. Types of Large Databases • Claims databases (e.g., Medicaid database) • Cooper et al. Major congenital malformations after first-trimester exposure to ACE inhibitors. N Engl J Med 2006: 354: 2443-51. • Electronic medical records (e.g., GPRD) • Jick SS. Pregancy outcomes after maternal exposure to fluconazole. Pharcmacotherapy 1999; 19:221-2. • Teratology information services • Levy et al. Pregnancy outcome following in utero exposure to bisphosphonates. Bone 2009; 44:428-30. • Population registries in Nordic countries • Wogelius et al. Maternal use of selective serotonin reuptake inhibitors and risk of congenital malformations. Epidemiology 2006: 17:701-4.

  25. Examples of Used Cohorts1: US Collaborative Perinatal Project (primary data collection) which enrolled over 50 000 women between 1959 and 1965, obtained detailed information on their pregnancies, and followed the children until age 7 (the overall size of the database is the major weakness) Example: Naeye RL. Causes of perinatal mortality in the US Collaborative Perinatal Project. JAMA 1977; 238:228-9. Medicaid Program, US (prescription data) 29 500pregnancies (the data quality and lack of records for the offspring are the major weaknesses) Example: Cooper et al. Major congenital malformations after first-trimester exposure to ACE inhibitors. N Engl J Med 2006: 354: 2443-51.

  26. Examples of Used Cohorts2: • North Jutland County Cohort • 230 000 pregnant women • Example:Wogelius et al. Maternal use of selective serotonin reuptake inhibitors and risk of congenital malformations. Epidemiology 2006: 17:701-4. • (the sample size is the major weakness) • The Danish National Birth Cohort • approx. 100,000 pregnant women • Example: Pedersen LH et al. Selective serotonin reuptake inhibitors in pregnancy and congenital malformations: population based cohort study. BMJ 2009;339:b3569.

  27. Examples of Used Cohorts3: • Motherisk Program, The Toronto Teratogen Information Service, Canada • Example: Levy et al. Pregnancy outcome following in utero exposure to bisphosphonates. Bone 2009; 44:428-30 • The Swedish Birth Registry • > 500 000 pregnancies • Example: Cleary BJ, Källén B.Early pregnancy azathioprine use and pregnancy outcomes. Birth Defects Res A Clin Mol Teratol 2009;85:647-54.

  28. Frequently Ignored Limitations of Prescription Databases and Registries The data collection methods are predetermined, and not controlled by research and sometimes impossible to validate. Misclassification exists in all data. The relatively large number of data may lead to data dredging and misleading post hoc analysis Random errors may be reduced, but systematic errors cannot be ruled out Basically most designs are cross-sectional studies

  29. Adult life Birth Conception Morbidity Spontaneous abortions Pre-natal diagnosis of birth defects Stillbirth Birth defects Preterm birth Low birth weight

  30. Selection Bias Inability to observe all outcomes in a cohort Spontaneous and induced fetal loss related to drug exposure induces selection bias If malformations are measured at birth, drugs related to survival may misleadingly appear to have/lack teratogenicity

  31. All pregnancies in Denmark 2007

  32. All birth defects in Denmark 2007

  33. Information Bias: Misclassification Errors in measurement of medication use: timing, adherence, “immortal person-time” Errors in ascertainment of malformations (10-50%); high specificity is more important than high sensitivity Errors in measurement of confounders results in residual confounding

  34. Confounding Unmeasured/unknown confounding: lack of data and randomization Residual confounding: due to misclassification, improper categorization of confounding variables Confounding by indication: disease for which a drug is prescribed is an independent risk factor for a given malformation

  35. What to Do about Bias? Quantify: sensitivity analysis of estimates under a range of assumption about bias direction and magnitude Correct: external adjustment Acknowledge and discuss

  36. The Future • Attention to: • Statistical power - we need more data – international collaboration • Integration of meta-analysis in order to solve some of the problems with statistical power. • Improve data quality • Secular trends in exposure (new drugs) • Better integration of epidemiology, biostatistics, pharmacology and biology

  37. Selected results – risk of spontaneous abortion NSAID exposure prevalence Adjusted OR* Abortion Live birth (95% CI) Time 0.07% 0.03% 7.0 (2.8-17.7) 1 week 2 week 3-4 weeks 5-6 weeks 7-12 weeks 0.12% 0.05% 3.0 (1.2-7.4) 0.33% 0.14% 4.4 (2.7-7.2) 0.45% 0.31% 2.7 (1.8-4.0) 0.52% 0.54% 1.3 (0.9-1.9) *adjusted for maternal age and mutually adjusted for use of NSAIDs

More Related