The development and widespread use of prenatal diagnostic techniques has allowed physicians to identify birth defects well before the expected date of delivery. Screening for neural tube defects using maternal serum alpha-fetoprotein levels and identification of Down syndrome through amniocentesis are now regular parts of obstetric practice. High-resolution ultrasound and fetal echocardiography permit the diagnosis of a wide range of anomalies from major defects such as skeletal dysplasias, diaphragmatic hernia, gastroschisis, and hypoplastic left heart to more subtle problems such as urinary tract dilations and choroid plexus cysts. 1–4 Advances in three-dimensional ultrasound and fetal magnetic resonance imaging are improving the ability to visualize conditions such as cleft lip, club foot, syndactyly, and facial dysmorphisms as well as the finer characteristics of major defects identified first by two-dimensional ultrasound. 5,6 As a result of these advances, women may now be offered a choice of options in the management of affected pregnancies, including elective termination. The exercise of these options has significant implications for epidemiologic studies of birth defects.
Estimates of the prevalence of neural tube defects indicate that, in some populations, up to 50% of all pregnancies with any neural tube defect, and up to 80% of those with anencephaly, may be electively terminated after prenatal diagnosis. 7–9 Because such terminations may occur outside the in-patient hospital setting, studies of neural tube defects that ascertain defects only among infants born in hospitals systematically exclude these pregnancies. A variety of factors may influence whether a congenital defect is diagnosed prenatally and how an affected pregnancy is managed. These factors include patients’ and practitioners’ access to and use of prenatal diagnostic techniques; the sensitivity, specificity, and predictive value of the diagnostic techniques for specific defects; the skill of the provider using these techniques; the nature and perceived severity of the defect diagnosed; and the decisions made by individual women concerning pregnancy management. If one or a combination of these variables is associated with the risk factor of interest, the result of an epidemiologic study may be biased if electively terminated pregnancies are not included.
In this paper, we explore, using examples from the literature, the potential effect that excluding prenatally diagnosed and electively terminated pregnancies has on case-control studies of birth defects. Although the studies cited here examine neural tube defects, the conclusions are applicable to studies of any defect for which prenatal diagnosis and elective termination are realistic options.
Subjects and Methods
In classic case-control studies, odds ratios (ORs) are calculated as AD/BC, where
A = the number of exposed pregnancies with the defect,
B = the number of exposed pregnancies without the defect,
C = the number of unexposed pregnancies with the defect, and
D = the number of unexposed pregnancies without the defect.
In the presence of prenatal diagnosis and elective termination among affected pregnancies, ORs are calculated as AS1D/ BCS2, where
S1 = the percentage of exposed pregnancies with the defect that are not terminated,
S2 = the percentage of unexposed pregnancies with the defect that are not terminated,
AS1 = the number of exposed pregnancies with the defect that are not terminated, and
CS2 = the number of unexposed pregnancies with the defect that are not terminated.
The bias in the OR due to the exclusion of prenatally diagnosed and electively terminated pregnancies is calculated as S1/ S2.
The exclusion of prenatally diagnosed defects among electively terminated pregnancies will bias the risk estimate in a case-control study only if prenatal diagnosis followed by elective termination is related to both the risk factor and the birth defect being studied in one of three possible ways: (1) the risk factor of interest is not related to whether an affected pregnancy is terminated after prenatal diagnosis (S1 =S2), (2) the presence of the risk factor is associated with an increased likelihood that an affected pregnancy is terminated after prenatal diagnosis (S1 <S2), and (3) the presence of the risk factor is associated with a decreased likelihood that an affected pregnancy is terminated after prenatal diagnosis (S1 >S2). We assessed these relations using data from published and theoretical studies of neural tube defects.
To assess the bias when S1 =S2, we used data from the Atlanta Birth Defects Case Control Study from 1968 through 1980 10 evaluating the relation between multivitamin use by pregnant women and their risk of having a pregnancy affected by anencephaly. Infants used as control subjects in this study did not have birth defects. To calculate the potential effect of prenatal diagnosis and elective termination on this relation, we assumed that 59% of pregnancies with anencephaly were electively terminated and that this percentage did not change with multivitamin use. This estimate was taken from a study of the effect of prenatal diagnosis on neural tube defect rates on the data of the Metropolitan Atlanta Congenital Defects Program, an active surveillance system for birth defects in the same population, from 1990 through 1991. 7
To assess the bias when S1 <S2, we also used data from the Metropolitan Atlanta Congenital Defects Program for 1990–1991 7 to evaluate the effect of prenatal diagnosis and elective termination on the relation between race and the risk of having a pregnancy affected by anencephaly. In these data, 64% of pregnancies with anencephaly among white women were terminated compared with 38% of such pregnancies among black women. Because only affected infants are ascertained by the Metropolitan Atlanta Congenital Defects Program, the data on affected pregnancies were applied to a theoretical case-control study (N = 99), in which infants used as control subjects had the same racial distribution as the Atlanta population in 1990 and were assumed not to have birth defects.
To assess the bias when S1 >S2, we again used data from the Atlanta Birth Defects Case Control Study from 1968 through 1980 11 evaluating the relation between maternal obesity and the risk of having a pregnancy affected by spina bifida. (The spina bifida-specific data were obtained from personal communication with Margaret L. Watkins, Centers for Disease Control and Prevention, 1999). Infants used as control subjects in this study did not have birth defects. To calculate the potential effect of prenatal diagnosis and elective termination on this relation, we assumed that 26% of pregnancies with spina bifida were electively terminated and that this percentage was not associated with maternal obesity. This estimate was taken from the data of the Metropolitan Atlanta Congenital Defects Program for 1990–1991. 7 We also assumed a 17% reduction in physicians’ ability to diagnosis spina bifida prenatally among obese women. This estimate was taken from a 1988 study of a population in Michigan. 12
We calculated 95% confidence intervals (CIs) around the ORs using Cornfield’s approximation as described by Fleiss, 13 unless an expected cell value was less than 5. In the latter instance, we calculated exact CIs using a method based on an algorithm and program by Mehta et al.14 The software used was Epi Info, version 6.04b. 15
Table 1 shows the nondifferential effect (S1 =S2) of prenatal diagnosis and elective termination on the relation between multivitamin use by pregnant women and their risk of having a pregnancy affected by anencephaly. 10 The 59% estimate of the rate of elective termination for anencephaly 7 was applied to the data for both women who did and those who did not take multivitamins. 1 This calculation yielded essentially the same OR (0.47 vs 0.46) regardless of whether prenatally diagnosed pregnancies are included in the case ascertainment. The 95% CI was wider when prenatally diagnosed cases were excluded.
Table 2 shows the differential effect (S1 <S2) of prenatal diagnosis and elective termination on the relation between race and the risk of anencephaly. 7 The 64% estimate of termination among pregnancies with anencephaly was applied to the data for white women, and the 38% estimate of termination among such pregnancies was applied to the data for black women. This difference resulted in the OR being biased toward the null (1.33 vs 2.30) when prenatally diagnosed pregnancies were not included in the case ascertainment, although with a slightly narrower 95% CI.
Similarly, Table 3 shows the differential effect (S1 >S2) of prenatal diagnosis and elective termination on the relation between maternal obesity and spina bifida. 11 (The spina bifida-specific data were obtained from personal communication with Margaret L. Watkins, Centers for Disease Control and Prevention, 1999). The 26% estimate of the rate of elective termination for spina bifida 7 was applied to the data for both obese and nonobese women, and the 17% estimate of the reduction in physicians’ ability to diagnosis spina bifida prenatally 12 was applied to the data for obese women. Using these estimates, we calculated that 89% of pregnancies with spina bifida among obese women were not terminated, compared with 72% among nonobese women. This difference resulted in the OR being biased away from the null (2.56 vs 2.02) when prenatally diagnosed pregnancies were not included in the case ascertainment, with a wider 95% CI.
These examples show that the exclusion of pregnancies electively terminated after prenatal diagnosis of a birth defect may have a substantial effect on the results of epidemiologic studies of these defects. Depending on how the risk factor being studied is related to prenatal diagnosis and elective termination, the study precision may be decreased or the OR may be biased either toward or away from the null. In addition, the magnitude of the bias will vary according to the population characteristics, the frequency with which prenatal diagnosis and elective termination are used, and the strength of the association of the risk factor with these procedures.
Velie and Shaw 16 have shown that, in California, women who elect to terminate a pregnancy after prenatal diagnosis of a neural tube defect differ with regard to race/ethnicity, age, education, and employment status from women who do not choose termination. Forrester and Merz 17,18 have shown that, in Hawaii, women who elect to terminate a pregnancy after prenatal diagnosis of Down syndrome, omphalocele, and gastroschisis differ with regard to race/ethnicity, age, urban vs rural residence, and whether they have had a previous pregnancy from women who do not choose termination. Studies in which one or more of these maternal factors are associated with the risk factor of interest may yield biased results if terminated pregnancies are not included. Indeed, Velie and Shaw 16 calculated that excluding electively terminated pregnancies from their study of neural tube defects would produce effect estimates for certain subgroups of women that were biased away from the null.
The frequency with which affected pregnancies are terminated after prenatal diagnosis varies according to the defect being studied. Termination rates after prenatal diagnosis have been estimated at more than 80% for Down syndrome, 50% for bilateral renal agenesis, 40% for omphalocele, and 25% for hypoplastic left heart in some populations. 9,17,18 Exclusion of these pregnancies would be expected to have a sizable effect on study results. In contrast, termination rates have been estimated at only 3% for tetralogy of Fallot and less than 1% for uncomplicated transposition of the great arteries, lesions for which surgical correction is well established. 2 Exclusion of these pregnancies would be expected to have little effect on most study results. Studies evaluating parental decisions to terminate pregnancies affected with birth defects indicate that severity of the anomaly and its associated prognosis are important determinants. 19–21
The situation becomes even more complex in studies in which infants with birth defects other than the one being studied are used as control subjects in an attempt to minimize recall bias in maternal interviews. 22 Such studies assume that the association between the risk factor and the various defects among control subjects, when taken in the aggregate, is different from that among the case subjects. The exclusion of prenatally diagnosed and electively terminated pregnancies from both the case and control groups would bias the OR not only by S1/ S2, but also by a corresponding factor reflecting the frequency of prenatal diagnosis and elective termination among the exposed and unexposed control pregnancies. Again, the resulting bias could be in either direction, or the several biases could potentially cancel each other out.
In an alternative scenario, prenatal diagnostic techniques may result in increased rates of diagnosis of some defects even among pregnancies that are not terminated. These additional diagnoses are often less severe manifestations of a defect than would be ascertained among symptomatic infants at birth. Obstructive renal defects are an example of such defects. 3,23,24 If the milder forms of these defects have a different etiology or different risk factors than the more severe forms, the inclusion of affected pregnancies in the case group may also influence the study results.
In summary, developments in medical technology have changed the way that birth defects are diagnosed and that affected pregnancies are managed. These developments have implications for the epidemiologic study of defects. Whenever possible, researchers should include prenatally diagnosed defects among electively terminated pregnancies in the case ascertainment to minimize the effect on their study results.
Unfortunately, information about these defects is not readily available. State-based abortion reporting generally does not include the reason for termination or whether a defect was diagnosed. 25 It has been documented that the frequency of elective termination is significantly underreported in surveys of women of reproductive age. 26 Direct access to medical records may be particularly problematic because of the sensitive nature of information about prenatal diagnosis and elective termination.
In the absence of specific data, it is imperative that researchers consider the potential for relations between prenatal diagnosis and elective termination of affected pregnancies and the risk factors and outcomes of interest, and how these relations may vary among populations, when designing studies of birth defects and reporting their findings. Only with these changes can the validity of epidemiologic studies be maintained in the face of changing technology.
1. Dallaire L, Michaud J, Melancon SB, Potier M, Lambert M, Mitchell G, Boisvert J. Prenatal diagnosis
of fetal anomalies during second trimester of pregnancy: their characterization and delineation of defects in pregnancies at risk. Prenat Diagn 1991; 11: 629–635.
2. Bull C. Current and potential impact of fetal diagnosis on prevalence and spectrum of serious congenital heart disease at term in the UK. Lancet 1999; 354: 1242–1247.
3. Noia G, Masini L, Caruso A, Perrelli L, Calisti A, Salvaggio E, Mancuso S. Prenatal diagnosis
of congenital uropathies. Fetal Ther 1989; 4 (suppl 1): 40–51.
4. Lescale KB, Eddleman KA, Chervenak FA. Prenatal diagnosis
of structural anomalies. Curr Opin Obstet Gynecol 1992; 4: 249–253.
5. Naylor CS, Platt LD. The role of three-dimensional ultrasonography in evaluating the developing fetus. MD Comput 1999; 16: 46–48.
6. Hubbard AM, Harty P. Prenatal magnetic resonance imaging of fetal anomalies. Semin Roentgenol 1999; 34: 41–47.
7. Cragan JD, Roberts HE, Edmonds LD, Khoury MJ, Kirby RS, Shaw GM, Velie EM, Merz RD, Forrester MB, Williamson RA, Krishnamurti DS, Stevenson RE, Dean JH. Surveillance for anencephaly and spina bifida and the impact of prenatal diagnosis
: United States, 1985–1994. MMWR CDC Surveill Summ 1995; 44: 1–13.
8. Forrester MB, Merz RD, Yoon PW. Impact of prenatal diagnosis
and elective termination
on the prevalence of selected birth defects in Hawaii. Am J Epidemiol 1998; 148: 1206–1211.
9. Game E, Bergman U. Benzodiazepine use in pregnancy and major malformation or oral clefts (Letter). BMJ 1999; 319: 918.
10. Mulinare J, Cordero JF, Erickson JD, Berry RJ. Periconceptional use of multivitamins and the occurrence of neural tube defects. JAMA 1988; 260: 3141–3145.
11. Watkins ML, Scanlon KS, Mulinare J, Khoury MJ. Is maternal obesity a risk factor for anencephaly and spina bifida? Epidemiology 1996; 7: 507–512.
12. Wolfe HM, Sokol RJ, Martier SM, Zador IE. Maternal obesity: a potential source of error in sonographic prenatal diagnosis
. Obstet Gynecol 1990; 76: 339–342.
13. Fleiss JL. Statistical Methods for Rates and Proportions. 2nd ed. New York: John Wiley and Sons, 1981; 71–75.
14. Mehta CR, Patel NR, Gray R. Computing an exact confidence interval for the common odds ratio
in several 2 × 2 contingency tables. J Am Stat Assoc 1985; 80: 969–973.
15. Dean AG, Dean JA, Coulombier D, Brendel KA, Smith DC, Burton AH, Dicker RC, Sullivan K, Fagan RF, Arner TG. Epi Info, Version 6: A Word Processing, Database, and Statistics Program for Epidemiology on Microcomputers. Atlanta: Centers for Disease Control and Prevention, 1994.
16. Velie EM, Shaw GM. Impact of prenatal diagnosis
and elective termination
on prevalence and risk estimates of neural tube defects in California, 1989–1991. Am J Epidemiol 1996; 144: 473–479.
17. Forrester MB, Merz RD. Prenatal diagnosis
and elective termination
of Down syndrome in a racially mixed population in Hawaii, 1987–1996. Prenat Diagn 1999; 19: 136–141.
18. Forrester MB, Merz RD. Impact of demographic factors on prenatal diagnosis
and elective termination
because of abdominal wall defects, Hawaii, 1986–1997. Fetal Diagn Ther 1999; 14: 206–211.
19. Pryde PG, Isada NB, Hallak M, Johnson MP, Odgers AE, Evans MI. Determinants of parental decision to abort or continue after non-aneuploid ultrasound-detected fetal abnormalities. Obstet Gynecol 1992; 80: 52–56.
20. Drugan A, Greb A, Johnson MP, Krivchenia EL, Uhlmann WR, Moghissi KS, Evans MI. Determinants of parental decisions to abort for chromosomal abnormalities. Prenat Diagn 1990; 10: 483–490.
21. Grevengood C, Shulman LP, Dungan JS, Martens P, Phillips OP, Emerson DS, Felker RE, Simpson JL, Elias S. Severity of abnormality influences decision to terminate pregnancies affected with fetal neural tube defects. Fetal Diagn Ther 1994; 9: 273–277.
22. Mills JL, Rhoads GG, Simpson JL, Cunningham GC, Conley MR, Lassman MR, Walden ME, Depp OR, Hoffman HJ, the National Institute of Child Health and Human Development Neural Tube Defects Study Group. The absence of a relation between the periconceptional use of vitamins and neural-tube defects. N Engl J Med 1989; 321: 430–435.
23. Livera LN, Brookfield DSK, Egginton JA, Hawnaur JM. Antenatal ultra- sonography to detect fetal renal abnormalities: a prospective screening programme. BMJ 1989; 298: 1421–1423.
24. Mandell J, Peters CA, Retik AB. Current concepts in the perinatal diagnosis and management of hydronephrosis. Urol Clin North Am 1990; 17: 247–262.
25. Anonymous. Abortion surveillance: preliminary analysis - - United States, 1996. MMWR 1998; 47: 1025–1028, 1035.
26. Fu H, Darroch JE, Henshaw SK, Kolb E. Measuring the extent of abortion underreporting in the 1995 National Survey of Family Growth. Fam Plann Perspect 1998; 30: 128–133, 138.