- Email: [email protected]
Potentially preventable excess mortality among higher-order multiples
ORIGINAL RESEARCH
Potentially Preventable Excess Mortality Among Higher-Order Multiples Hamisu M. Salihu, MD, PhD, Muktar H. Aliyu, MD, MPH, Dwight J. Rouse, MD, MSPH, Russell S. Kirby, MS, PhD, and Greg R. Alexander, MPH, ScD OBJECTIVE: To estimate the level of potentially preventable excess mortality achievable by avoiding the creation of higher-order multiple gestation with assisted reproductive technologies. METHODS: This was a retrospective cohort study of multiple pregnancies delivered in the United States between 1995 and 1997 involving 304,466 twins, 16,068 triplets, 1448 quadruplets, and 180 quintuplets. We used the generalized estimating equation framework to compute adjusted relative risks for combined perinatal and infant mortality (early mortality). We then calculated potentially preventable excess mortality among higher-order gestations, using twins and triplets sequentially as the referent category. RESULTS: Early mortality increased significantly with each additional fetus in a dose-dependent fashion (P < .001), corresponding to relative risks (95% confidence interval) of 2.4 (2.2, 2.6) for triplets, 3.3 (2.5, 4.4) for quadruplets, and 10.3 (5.0, 21.4) for quintuplets. The creation of twin rather than quadruplet pregnancies would be associated with a substantially higher level of preventable excess mortality (70%) than the creation of triplet pregnancies (28%). By contrast, limiting quintuplets to twins or triplets did not exhibit a similar level of difference (89% versus 75%, respectively). CONCLUSION: Our findings support the need for regulating the number of transferred embryos that result in quadruplet and quintuplet pregnancies. (Obstet Gynecol 2003; 102:679 – 84. © 2003 by The American College of Obstetricians and Gynecologists.)
There has been a dramatic rise in the incidence of multiple births in the United States over the last two decades. The frequency of twin births has doubled; that of triplets has increased six-fold, of quadruplets almost 12-fold, and of quintuplets six-fold.1 Two phenomena account for most of this rise: increased maternal age From the Departments of Maternal and Child Health, Epidemiology, and Obstetrics and Gynecology, University of Alabama at Birmingham, Birmingham, Alabama.
secondary to delayed childbearing, and the use of assisted reproductive technology.2,3 The burden of morbidity and mortality seems to increase substantially with each additional fetus in a multiple gestation. It is known that 90% of triplets are born preterm, and triplets and other higher-order births are 12 times as likely to die during the first year of life compared with singleton infants.4 As a result, there has been growing debate on the need to prevent higher-order multiple pregnancies.5,6 However, surrounding this issue is the lack of population-level information on the potential benefits (in terms of preventable mortality) of the avoidance of higher-order multiple gestation. Prevention of higher-order multiple gestation is feasible chiefly through two methods: transfer of fewer embryos7 and fetal reduction. A third possibility for reducing higher-order multiples is through less intensive ovarian stimulation.8 We do not address fetal reduction in this article, because it is not a primary prevention procedure, involves risks to the ongoing pregnancy,5 and is ethically controversial. Less intensive ovarian stimulation might reduce the occurrence of higher-order multiples, though only to a limited degree and at the expense of overall pregnancy rates.8 Our analysis is therefore based on the possibility of primary prevention by limiting the number of embryos transferred during assisted reproductive technology cycles. Although limiting the number of transferred embryos might result in lower pregnancy success rates, recent studies have demonstrated that when embryos are transferred at the blastocyst stage, the pregnancy success rates are very high.9,10 For instance, the transfer of only two such embryos results in a clinical pregnancy in up to 60% of women.9,10 Based on these high pregnancy success rates and the speculation that the procedure is likely to improve with more experience, we estimated the level of potentially preventable deaths that could occur by limiting the number of embryos transferred to either two or three.
VOL. 102, NO. 4, OCTOBER 2003 © 2003 by The American College of Obstetricians and Gynecologists. Published by Elsevier.
0029-7844/03/$30.00 doi:10.1016/S0029-7844(03)00768-3
679
Table 1. Proportion (Percentages) of Selected Maternal Sociodemographic Characteristics by Plurality
Age (y) ⬍35 ⱖ35 Parity Nullipara Multipara Missing data Education (y) ⬍12 ⱖ12 Missing data Marital status Married Unmarried Missing data Race White Nonwhite Prenatal care Adequate Inadequate Smoking Yes No Missing data Alcohol Yes No Missing data
Twins (n ⫽ 152,233*)
Triplets (n ⫽ 5356*)
Quadruplets (n ⫽ 362*)
Quintuplets (n ⫽ 36*)
82.3 17.7
72.6 27.4
77.4 22.6
77.8 22.2
⬍.001
34.3 65.1 0.6
51.1 48.0 0.9
54.4 45.6 0.0
44.4 55.6 0.0
⬍.001
16.4 82.0 1.6
5.0 93.2 1.8
3.0 95.6 1.4
5.6 88.9 5.4
⬍.001
71.7 27.9 0.4
89.9 9.4 0.7
95.9 3.0 1.1
97.2 0.0 2.8
⬍.001
78.8 21.2
89.4 10.6
94.8 5.2
91.7 8.3
⬍.001
70.1 29.9
80.8 19.2
78.7 21.3
77.8 22.2
⬍.001
9.7 70.1 20.2
3.3 77.5 19.2
1.1 82.0 16.9
2.8 77.8 19.4
⬍.001
1.0 84.4 14.6
0.3 87.1 12.6
0.0 91.4 8.6
0.0 89.0 11.0
⬍.001
P
* n denotes the number of mothers and not the individual infants.
MATERIALS AND METHODS Data for this analysis were from the “Matched Multiple Birth File” assembled by the National Center for Health Statistics11 covering the period 1995–1997, the most recent for which matched and linked (birth linked to death) data for multiple deliveries in the United States were available. The file contains individual records of live births and fetal deaths involving multiple deliveries. Although the public-access version contains matched and linked information for twins and triplets only, the dataset that was made available to us upon request to the National Center for Health Statistics was matched and linked for higher-order pregnancies as well, with personal identifiers expunged. The completeness of this file is excellent (99%), and the procedures for quality control of the data are explained in detail elsewhere.12 We compared the following maternal sociodemographic characteristics among twin, triplet, quadruplet, and quintuplet pregnancies: maternal age, parity, race or ethnicity, marital status, education, level of prenatal care utilization, and smoking and drinking alcohol during
680
Salihu et al
Early Mortality Among Multiples
pregnancy. Parity refers to the total number of pregnancies reaching at least 20 weeks that the mother had experienced and was classified into nullipara (0) and multipara (1 or more). Adequacy of prenatal care was determined with the revised graduated index algorithm13,14 and was categorized into adequate and inadequate prenatal care utilization. This index of prenatal care has been found to be more accurate than many others, especially in describing the level of prenatal care utilization among groups that are at high risk and therefore exposed to intense care (eg, multiple pregnancies).15 In this study, inadequate prenatal care utilization refers to women who either had missing prenatal care information, had prenatal care but at a suboptimal level, or had no prenatal care at all. The accuracy of all these aforementioned variables on the birth certificate has been validated in previous studies and found to be accurately reported.16,17 For this study, the main birth outcome of interest was early mortality, the components of which were the following: stillborn fetus (inutero demise at 20 or higher weeks’ gestation), neonatal mortality (death of the new-
OBSTETRICS & GYNECOLOGY
Figure 1. Crude mortality rates among twins, triplets, quadruplets, and quintuplets in the United States, 1995–1997. Salihu. Early Mortality Among Multiples. Obstet Gynecol 2003.
born through the 27th day of life), and infant mortality (death within the first year of life). The fact that siblings from the same pregnancy are subjected to similar prevailing conditions in the womb and are more or less equally affected by the same attributes that characterize the mother makes the application of conventional statistical techniques inappropriate, because they treat observations as though they are independent. To avoid this bias, we applied the generalized estimating equation18 to our data to adjust for dependence. We used PROC GENMOD 8.2 software (SAS Institute, Cary, NC) to generate random effect estimates that captured sibling correlations. All tests of hypothesis were two-tailed, with a Type 1 error rate fixed at 5%. We calculated the preventive fraction, namely, the level of excess early mortality that could be prevented if higher-order multiples (quadruplets and quintuplets) were avoided and lower-order births (triplets and twins) created in their stead. The following formula was used19:
PF ⫽ 关共RR ⫺ 1兲/RR兴 ⫻ 100,
where PF ⫽ preventable fraction (or excess preventable mortality) and RR ⫽ adjusted relative risk (odds ratios). We computed two forms of PF. In one, we assumed a scenario in which higher-order multiples were replaced by twins; in the other, the assumed scenario was that they were replaced by triplets. Finally, we compared the relative benefits of both scenarios (quadruplets-totwins versus quadruplets-to-triplets and quintuplets-totwins versus quintuplets-to-triplets) in terms of the potentially preventable excess mortality. This study was approved by the Institutional Review Board of the University of Alabama at Birmingham (protocol number: X030527005). RESULTS The data set contains a total of 326,231 individual multiple deliveries in the United States for the period
Table 2. Absolute Counts of Early Mortality Among Twins, Triplets, and Higher-Order Multiples in the United States, 1995–1997 Stillbirth Neonatal death Infant death Early mortality
Twins
Triplets
Quadruplets
Quintuplets
5624 (1.8) 7375 (2.4) 9005 (3.0) 14,629 (4.8)
378 (2.4) 887 (5.5) 1002 (6.2) 1380 (8.6)
53 (3.7) 94 (6.5) 104 (7.2) 157 (10.8)
10 (5.6) 41 (22.8) 42 (23.3) 52 (28.9)
Numbers in parentheses are percentages.
VOL. 102, NO. 4, OCTOBER 2003
Salihu et al
Early Mortality Among Multiples
681
Table 3. Mortality Estimates by Plurality Groups, Obtained With the Generalized Estimating Equation Framework With Twin Category as the Referent Group Triplets Stillborn fetus Neonatal mortality Infant mortality Early mortality
Quadruplets
Quintuplets
AOR*
95% CI
AOR*
95% CI
AOR*
95% CI
1.8 2.8 2.7 2.4
1.6, 2.1 1.1, 3.1 2.4, 2.9 2.2, 2.6
3.4 3.4 3.2 3.3
2.1, 5.4 2.4, 5.0 2.3, 4.5 2.5, 4.4
3.9 15.2 13.3 10.3
1.1, 14.6 6.9, 33.9 6.1, 29.0 5.0, 21.4
AOR ⫽ adjusted odds ratio; CI ⫽ confidence interval. P for trend ⬍.001 for all the four mortality indices. * Adjusted estimates were generated by taking into account the confounding effects of maternal age, race, education, parity, adequacy of prenatal care, marital status, smoking, and alcohol ingestion during pregnancy.
1995–1997. Of these, twins, triplets, quadruplets, and quintuplets comprised 94.4%, 5.1%, 0.5%, and 0.07% respectively. Complete matching and linkage was achieved for 322,162 (98.8%) of the records. These represent 304,466 twins, 16,068 triplets, 1448 quadruplets, and 180 quintuplets. The distribution of selected sociodemographic characteristics of the mothers is shown in Table 1. Figure 1 displays crude mortality rates for twins, triplets, quadruplets, and quintuplets; the absolute counts of early mortality and its components are shown in Table 2. Table 3 summarizes the adjusted estimates for early mortality and its components with twins used as the referent category. Overall, triplets, quadruplets, and quintuplets had higher risks of mortality than twins. Generally, the twins versus higher-multiple disparity in mortality tended to be lower for stillborn fetuses than for neonatal and infant mortality. The most pronounced discrepancy was for quintuplets, among whom the risk for mortality was 4 –15 times that of twins. In contrast to the modest differences in risk levels observed for triplets and quadruplets, quintuplets demonstrated a wide gap between intrauterine and extrauterine mortality risks (Table 3). Table 4 shows the results for the various mortality indices when triplets were used as the referent category and comparisons made with higher-order births, and Table 4. Mortality Estimates by Plurality Groups, Obtained With the Generalized Estimating Equation Framework With Triplet Category as Referent Group Quadruplets Stillborn fetus Neonatal mortality Infant mortality Early mortality
Quintuplets
AOR*
95% CI
AOR*
95% CI
2.2 1.2 1.2 1.4
1.3, 3.6 0.8, 1.8 0.8, 1.7 1.1, 1.9
2.5 5.3 4.9 4.4
0.6, 10.0 2.2, 12.7 2.1, 11.3 2.0, 9.5
Abbreviations as in Table 3. * Adjusted estimates were generated by taking into account the confounding effects of maternal age, race, education, parity, adequacy of prenatal care, marital status, smoking, and alcohol ingestion during pregnancy.
682
Salihu et al
Early Mortality Among Multiples
Table 5 contains the proportion of excess mortality that could be prevented by replacing higher-order multiple gestations with triplets and twins through the restriction of the number of transferred embryos. The proportion of excess mortality that could be prevented among triplets by replacing them with twins ranged from 45% to 64%, with a lower magnitude of preventive fraction for intrauterine demise and higher values for extrauterine death. Among quadruplets, the level of preventable deaths that could be achieved by replacing them with twins was around 70%, with practically no variation across mortality types. The most remarkable benefit of limitation was clearly in the case of quintuplets. If they were replaced by twins, the avoidable excess mortality would be in the range of 74 –93%. The right side of Table 5 features the proportion of excess mortality preventable by replacing quadruplets and quintuplets with triplets instead of twins. The level of preventable deaths was still higher for quintuplets (60 – 81%) compared with that for quadruplets (15–54%). DISCUSSION This study provides risk estimates for early mortality as well as the magnitude of excess deaths potentially preventable by avoiding higher-order multiple gestation through the restriction of the number of transferred embryos during assisted reproductive technology cycles. We found that the likelihood of stillborn fetus and neonatal and infant mortality increased in a dose-dependent fashion with each additional fetus. In general, there are very limited data on mortality outcomes among higher-order births beyond triplets and quadruplets.20,21 An important limitation in these previous studies is the lack of adjustment for several potentially confounding characteristics, which were considered and controlled for in our study. Thus, it is reasonable to assume that the risk estimates reported in this study provide the most comprehensive and minimally biased risk estimates for stillborn fetus and neonatal and infant mortality among twins, triplets, quadruplets, and quintuplets to date in the
OBSTETRICS & GYNECOLOGY
Table 5. Preventive Fraction for Stillborn Fetuses, Neonatal Mortality, Perinatal Mortality, and Infant Mortality by Plurality Groups Preventive fraction by limiting to triplets
Preventive fraction by limiting to twins Stillborn fetus Neonatal mortality Infant mortality Early mortality
Triplets
Quadruplets
Quintuplets
Quadruplets
Quintuplets
169 (44.8) 570 (64.3) 625 (62.4) 794 (57.5)
37 (70.4) 67 (70.9) 72 (69.1) 109 (69.4)
7 (74.2) 38 (93.4) 39 (92.5) 46 (88.5)
28 (53.7) 15 (16.0) 16 (15.3) 44 (28.0)
6 (60.0) 33 (81.2) 33 (79.5) 39 (75.0)
Data are presented as n (%).
United States. To that extent, the results of our study should be useful for counseling women considering infertility treatment options. An important aspect of this study is the calculated preventive fraction for early mortality. These are adjusted estimates for the numbers of deaths that could be prevented by avoiding the creation of higher-order multiple gestations and replacing them with either twins or triplets. This information should help to inform the current debate regarding the need for averting higherorder multiple gestations by restricting the number of transferred embryos.5 A previous study has used data from the United Kingdom to examine the cost effectiveness of two- versus threeembryo transfers and found the benefits to be comparable.22 A cost– benefit or cost– utility analysis of this nature, especially, for higher multiples in the United States, though clearly beyond the scope of this article, is indicated. Such a study could characterize the costs and benefits of preventing higher-order births through the restriction of the number of transferred embryos, one potential disadvantage of which is a reduced pregnancy success rate. However, the ability to culture embryos to the blastocyst stage23 now permits the transfer of as few as two embryos with pregnancy occurring in up to 60% of women.9,10 Assuming this pregnancy rate to be fixed, then to avoid higher-order multiples, the number of transferred blastocysts could be limited to a maximum of two to three with substantially high pregnancy success rates. Currently, the American Society for Reproductive Medicine recommends the transfer of two good quality embryos for patients with favorable prognosis, three for above-average, four for average, and no more than five for below-average cases.24 In other industrialized countries, the maximum number of transferable embryos is lower than that recommended by the American Society for Reproductive Medicine. The British Fertility Society recommends a maximum of two embryos.25 The European Society for Human Reproduction and Embryology26 advocates either two or three, depending on female age. Other fertility experts have strongly suggested
VOL. 102, NO. 4, OCTOBER 2003
that we should be moving toward the transfer of only one embryo.27 Hence, our recommendation of a maximum of two to three blastocysts based on the analysis above seems reasonable. A clear limitation in this study is our lack of information on assisted reproduction procedures. This information is not available in the birth record of infants born in the United States. As a result, we were unable to conduct separate analyses for artificially and naturally conceived multiple gestations, However, currently almost all quadruplets and quintuplets are the results of assisted reproduction technology,20 and because the focus of our analysis was mainly on these plurality groups it is unlikely that information on whether these pregnancies were natural or artificial would have influenced our findings. On the other hand, triplets and twins in this study were conceived either spontaneously or through assisted reproductive technology. It is therefore likely that the risk estimates obtained with twins and triplets used as referent categories could have been impacted by this heterogeneous composition. Another limitation of this study is the lack of information on those multiples that underwent reduction procedures to lower-order multiples. In other words, it is possible that some of the triplets and twins we used as referents were actually the results of reduction procedures of previous higher-order multiples. Currently, the available information suggests that survivors of reduction procedures face either higher-than-expected risks of morbidity or mortality, or, at best, similar risks as twins and triplets that did not result from reduction procedures.28,29 Therefore, it is likely that our computed values for preventive fractions represent conservative estimates. In summary, our findings, based on comprehensive US data, support that there would be substantial, quantifiable benefit in preventing the occurrence of higherorder multiple gestations and provide a sound rationale for limiting the number of embryos transferred during assisted reproduction procedures.
Salihu et al
Early Mortality Among Multiples
683
REFERENCES 1. Kiely JL, Kiely M. Epidemiologic trends in multiple births in the United States, 1997-1998. Twin Res 2001;14:131–3. 2. Blickstein I, Goldman RD, Mazkereth R. Incidence and birth weight characteristics of twins born to mothers aged 40 years or more compared with 35-39 years old mothers: A population study. J Perinat Med 2001;29:128–32. 3. Ventura SJ. Trends and variations in first births to older women, 1970-86. National Center for Health Statistics. Vital Health Stat 1989;47:1–27. 4. Martin JA, Park MM. Trends in twin and triplet births: 1980-97. Natl Vital Stat Rep 1999;47:1–6. 5. Evans MI, Kramer RL, Yaron Y, Drugan A, Johnson MP. What are the ethical and technical problems associated with multifetal pregnancy reduction? Clin Obstet Gynecol 1998;41:47–54. 6. Torok O, Lapinski R, Salafia CM, Bernasko J, Berkowitz RL. Multifetal pregnancy reduction is not associated with an increased risk of intrauterine growth restriction, except for very-high-order multiples. Am J Obstet Gynecol 1998; 179:221–5. 7. Templeton A, Morris JK. Reducing the risk of multiple births by transfer of two embryos after in vitro fertilization. N Engl J Med 1998;339:573–7. 8. Gleicher N, Oleske D, Tur-Kaspa I, Vidali A, Karande V. Reducing the risk of high order multiple pregnancy after ovarian stimulation with gonadotropins. N Engl J Med 2000;343:2–7. 9. Gardner DK, Vella P, Lane M, Wagley L, Schlenker T, Schoolcraft WB. Culture and transfer of human blastocysts increases implantation rates and reduces the need for multiple embryo transfers. Fertil Steril 1998;69:84–8. 10. Gardner D, Schoolcraft WB, Wagley L, Schlenker T, Stevens J, Hesla J. A prospective randomized trial of blastocyst culture and transfer in in-vitro fertilization. Hum Reprod 1998;13:3434–40. 11. Martin J, Curtin S, Saulnier M, Mousavi J. Development of the matched multiple birth file. In: 1995-1997 matched multiple birth dataset. NCHS CD-ROM series 21, no. 12. Hyattsville, Maryland: National Center for Health Statistics, 2000. 12. National Center for Health Statistics. 1995-1997 linked birth/infant death data set. Vital statistics of the United States: Quality control procedures. Hyattsville, Maryland: US Department of Health and Human Services, Centers for Disease Control and Prevention, 17. 13. Alexander GR, Kotelchuck M. Quantifying the adequacy of prenatal care: A comparison of indices. Public Health Rep 1996;3:408–18. 14. Alexander GR, Cornely DA. Prenatal care utilization: Its measurement and relationship to pregnancy outcome. Am J Prev Med 1987;3:243–53. 15. Kogan MD, Martin JA, Alexander GR, Kotelchuck M, Ventura SJ, Frigoletto FD. The changing pattern of prena-
684
Salihu et al
Early Mortality Among Multiples
16.
17.
18. 19.
20. 21.
22.
23.
24.
25. 26.
27.
28.
29.
tal care utilization in the United States, 1981-1995, using different prenatal care indices. JAMA 1998;279:1623–8. Buescher PA, Taylor KP, Davis MH, Bowling JM. The quality of the new birth certificate data: A validation study in North Carolina. Am J Public Health 1993;83:1163–5. DiGiuseppe DL, Aron DC, Ranbom L, Harper DL, Rosenthal GE. Reliability of birth certificate data: A multihospital comparison to medical records information. Matern Child Health J 2002;16:169–79. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986;42:121–30. Hennekens CH, Buring JE. Measures of disease frequency and association. In: Mayrent SL, ed. Epidemiology in medicine, 1st ed. Boston: Little, Brown and Company, 1987:88. Francois K, Alperin A, Elliott JP. Outcomes of quintuplet pregnancies. J Reprod Med 2001;46:1047–51. Skrablin S, Kuvacic I, Pavicic D, Kalafatic D, Goluza T. Maternal neonatal outcome in quadruplet and quintuplet versus triplet gestations. Eur J Obstet Gynecol 2000;88: 147–52. Keith LG, Blickstein I, Oleszczuk JJ, Keith DM, eds. Triplet pregnancies and their consequences. London: The Parthenon Publishing Group, 2002. Gardner DK, Lane M. Culture of viable human blastocysts in defined sequential serum-free media. Hum Reprod 1998;13:148–59. American Society for Reproductive Medicine. Guidelines on number of embryos transferred: A Practice Committee report: a committee opinion. Birmingham, Alabama: American Society for Reproductive Medicine, 1999. Murdoch AP. How many embryos should be transferred? Hum Reprod 1998;13:2666–70. ESHRE Capri Workshop. Infertility revisited: The state of the art today and tomorrow. Hum Reprod 1996;11: 1779–807. Coetsier T, Dhont M. Avoiding multiple pregnancies in IVF: Who’s afraid of single embryo transfer? Hum Reprod 1998;13:2663–4. Papageorghiou AT, Liao AW, Skentou C, Sebire NJ, Nicolaides KH. Trichorionic triplet pregnancies at 10-14 weeks: Outcome after embryo reduction compared to expectant management. J Matern Fetal Neonatal Med 2002;11: 307–12. Jirsova S, Mardesic T, Muller P, Huttelova R, Zvarova J, Jirkovsky M. Multi-fetal pregnancy reduction does not influence perinatal results in twin pregnancies. Twin Res 2001;4:422–5.
Address reprint requests to: Hamisu Salihu, MD, PhD, University of Alabama at Birmingham, Department of Maternal and Child Health, 1665 University Boulevard, Room 320, Birmingham, Alabama 35294; E-mail: [email protected]. Received April 14, 2003. Received in revised form June 15, 2003. Accepted June 26, 2003.
OBSTETRICS & GYNECOLOGY
Potentially Preventable Excess Mortality Among Higher-Order Multiples Hamisu M. Salihu, MD, PhD, Muktar H. Aliyu, MD, MPH, Dwight J. Rouse, MD, MSPH, Russell S. Kirby, MS, PhD, and Greg R. Alexander, MPH, ScD OBJECTIVE: To estimate the level of potentially preventable excess mortality achievable by avoiding the creation of higher-order multiple gestation with assisted reproductive technologies. METHODS: This was a retrospective cohort study of multiple pregnancies delivered in the United States between 1995 and 1997 involving 304,466 twins, 16,068 triplets, 1448 quadruplets, and 180 quintuplets. We used the generalized estimating equation framework to compute adjusted relative risks for combined perinatal and infant mortality (early mortality). We then calculated potentially preventable excess mortality among higher-order gestations, using twins and triplets sequentially as the referent category. RESULTS: Early mortality increased significantly with each additional fetus in a dose-dependent fashion (P < .001), corresponding to relative risks (95% confidence interval) of 2.4 (2.2, 2.6) for triplets, 3.3 (2.5, 4.4) for quadruplets, and 10.3 (5.0, 21.4) for quintuplets. The creation of twin rather than quadruplet pregnancies would be associated with a substantially higher level of preventable excess mortality (70%) than the creation of triplet pregnancies (28%). By contrast, limiting quintuplets to twins or triplets did not exhibit a similar level of difference (89% versus 75%, respectively). CONCLUSION: Our findings support the need for regulating the number of transferred embryos that result in quadruplet and quintuplet pregnancies. (Obstet Gynecol 2003; 102:679 – 84. © 2003 by The American College of Obstetricians and Gynecologists.)
There has been a dramatic rise in the incidence of multiple births in the United States over the last two decades. The frequency of twin births has doubled; that of triplets has increased six-fold, of quadruplets almost 12-fold, and of quintuplets six-fold.1 Two phenomena account for most of this rise: increased maternal age From the Departments of Maternal and Child Health, Epidemiology, and Obstetrics and Gynecology, University of Alabama at Birmingham, Birmingham, Alabama.
secondary to delayed childbearing, and the use of assisted reproductive technology.2,3 The burden of morbidity and mortality seems to increase substantially with each additional fetus in a multiple gestation. It is known that 90% of triplets are born preterm, and triplets and other higher-order births are 12 times as likely to die during the first year of life compared with singleton infants.4 As a result, there has been growing debate on the need to prevent higher-order multiple pregnancies.5,6 However, surrounding this issue is the lack of population-level information on the potential benefits (in terms of preventable mortality) of the avoidance of higher-order multiple gestation. Prevention of higher-order multiple gestation is feasible chiefly through two methods: transfer of fewer embryos7 and fetal reduction. A third possibility for reducing higher-order multiples is through less intensive ovarian stimulation.8 We do not address fetal reduction in this article, because it is not a primary prevention procedure, involves risks to the ongoing pregnancy,5 and is ethically controversial. Less intensive ovarian stimulation might reduce the occurrence of higher-order multiples, though only to a limited degree and at the expense of overall pregnancy rates.8 Our analysis is therefore based on the possibility of primary prevention by limiting the number of embryos transferred during assisted reproductive technology cycles. Although limiting the number of transferred embryos might result in lower pregnancy success rates, recent studies have demonstrated that when embryos are transferred at the blastocyst stage, the pregnancy success rates are very high.9,10 For instance, the transfer of only two such embryos results in a clinical pregnancy in up to 60% of women.9,10 Based on these high pregnancy success rates and the speculation that the procedure is likely to improve with more experience, we estimated the level of potentially preventable deaths that could occur by limiting the number of embryos transferred to either two or three.
VOL. 102, NO. 4, OCTOBER 2003 © 2003 by The American College of Obstetricians and Gynecologists. Published by Elsevier.
0029-7844/03/$30.00 doi:10.1016/S0029-7844(03)00768-3
679
Table 1. Proportion (Percentages) of Selected Maternal Sociodemographic Characteristics by Plurality
Age (y) ⬍35 ⱖ35 Parity Nullipara Multipara Missing data Education (y) ⬍12 ⱖ12 Missing data Marital status Married Unmarried Missing data Race White Nonwhite Prenatal care Adequate Inadequate Smoking Yes No Missing data Alcohol Yes No Missing data
Twins (n ⫽ 152,233*)
Triplets (n ⫽ 5356*)
Quadruplets (n ⫽ 362*)
Quintuplets (n ⫽ 36*)
82.3 17.7
72.6 27.4
77.4 22.6
77.8 22.2
⬍.001
34.3 65.1 0.6
51.1 48.0 0.9
54.4 45.6 0.0
44.4 55.6 0.0
⬍.001
16.4 82.0 1.6
5.0 93.2 1.8
3.0 95.6 1.4
5.6 88.9 5.4
⬍.001
71.7 27.9 0.4
89.9 9.4 0.7
95.9 3.0 1.1
97.2 0.0 2.8
⬍.001
78.8 21.2
89.4 10.6
94.8 5.2
91.7 8.3
⬍.001
70.1 29.9
80.8 19.2
78.7 21.3
77.8 22.2
⬍.001
9.7 70.1 20.2
3.3 77.5 19.2
1.1 82.0 16.9
2.8 77.8 19.4
⬍.001
1.0 84.4 14.6
0.3 87.1 12.6
0.0 91.4 8.6
0.0 89.0 11.0
⬍.001
P
* n denotes the number of mothers and not the individual infants.
MATERIALS AND METHODS Data for this analysis were from the “Matched Multiple Birth File” assembled by the National Center for Health Statistics11 covering the period 1995–1997, the most recent for which matched and linked (birth linked to death) data for multiple deliveries in the United States were available. The file contains individual records of live births and fetal deaths involving multiple deliveries. Although the public-access version contains matched and linked information for twins and triplets only, the dataset that was made available to us upon request to the National Center for Health Statistics was matched and linked for higher-order pregnancies as well, with personal identifiers expunged. The completeness of this file is excellent (99%), and the procedures for quality control of the data are explained in detail elsewhere.12 We compared the following maternal sociodemographic characteristics among twin, triplet, quadruplet, and quintuplet pregnancies: maternal age, parity, race or ethnicity, marital status, education, level of prenatal care utilization, and smoking and drinking alcohol during
680
Salihu et al
Early Mortality Among Multiples
pregnancy. Parity refers to the total number of pregnancies reaching at least 20 weeks that the mother had experienced and was classified into nullipara (0) and multipara (1 or more). Adequacy of prenatal care was determined with the revised graduated index algorithm13,14 and was categorized into adequate and inadequate prenatal care utilization. This index of prenatal care has been found to be more accurate than many others, especially in describing the level of prenatal care utilization among groups that are at high risk and therefore exposed to intense care (eg, multiple pregnancies).15 In this study, inadequate prenatal care utilization refers to women who either had missing prenatal care information, had prenatal care but at a suboptimal level, or had no prenatal care at all. The accuracy of all these aforementioned variables on the birth certificate has been validated in previous studies and found to be accurately reported.16,17 For this study, the main birth outcome of interest was early mortality, the components of which were the following: stillborn fetus (inutero demise at 20 or higher weeks’ gestation), neonatal mortality (death of the new-
OBSTETRICS & GYNECOLOGY
Figure 1. Crude mortality rates among twins, triplets, quadruplets, and quintuplets in the United States, 1995–1997. Salihu. Early Mortality Among Multiples. Obstet Gynecol 2003.
born through the 27th day of life), and infant mortality (death within the first year of life). The fact that siblings from the same pregnancy are subjected to similar prevailing conditions in the womb and are more or less equally affected by the same attributes that characterize the mother makes the application of conventional statistical techniques inappropriate, because they treat observations as though they are independent. To avoid this bias, we applied the generalized estimating equation18 to our data to adjust for dependence. We used PROC GENMOD 8.2 software (SAS Institute, Cary, NC) to generate random effect estimates that captured sibling correlations. All tests of hypothesis were two-tailed, with a Type 1 error rate fixed at 5%. We calculated the preventive fraction, namely, the level of excess early mortality that could be prevented if higher-order multiples (quadruplets and quintuplets) were avoided and lower-order births (triplets and twins) created in their stead. The following formula was used19:
PF ⫽ 关共RR ⫺ 1兲/RR兴 ⫻ 100,
where PF ⫽ preventable fraction (or excess preventable mortality) and RR ⫽ adjusted relative risk (odds ratios). We computed two forms of PF. In one, we assumed a scenario in which higher-order multiples were replaced by twins; in the other, the assumed scenario was that they were replaced by triplets. Finally, we compared the relative benefits of both scenarios (quadruplets-totwins versus quadruplets-to-triplets and quintuplets-totwins versus quintuplets-to-triplets) in terms of the potentially preventable excess mortality. This study was approved by the Institutional Review Board of the University of Alabama at Birmingham (protocol number: X030527005). RESULTS The data set contains a total of 326,231 individual multiple deliveries in the United States for the period
Table 2. Absolute Counts of Early Mortality Among Twins, Triplets, and Higher-Order Multiples in the United States, 1995–1997 Stillbirth Neonatal death Infant death Early mortality
Twins
Triplets
Quadruplets
Quintuplets
5624 (1.8) 7375 (2.4) 9005 (3.0) 14,629 (4.8)
378 (2.4) 887 (5.5) 1002 (6.2) 1380 (8.6)
53 (3.7) 94 (6.5) 104 (7.2) 157 (10.8)
10 (5.6) 41 (22.8) 42 (23.3) 52 (28.9)
Numbers in parentheses are percentages.
VOL. 102, NO. 4, OCTOBER 2003
Salihu et al
Early Mortality Among Multiples
681
Table 3. Mortality Estimates by Plurality Groups, Obtained With the Generalized Estimating Equation Framework With Twin Category as the Referent Group Triplets Stillborn fetus Neonatal mortality Infant mortality Early mortality
Quadruplets
Quintuplets
AOR*
95% CI
AOR*
95% CI
AOR*
95% CI
1.8 2.8 2.7 2.4
1.6, 2.1 1.1, 3.1 2.4, 2.9 2.2, 2.6
3.4 3.4 3.2 3.3
2.1, 5.4 2.4, 5.0 2.3, 4.5 2.5, 4.4
3.9 15.2 13.3 10.3
1.1, 14.6 6.9, 33.9 6.1, 29.0 5.0, 21.4
AOR ⫽ adjusted odds ratio; CI ⫽ confidence interval. P for trend ⬍.001 for all the four mortality indices. * Adjusted estimates were generated by taking into account the confounding effects of maternal age, race, education, parity, adequacy of prenatal care, marital status, smoking, and alcohol ingestion during pregnancy.
1995–1997. Of these, twins, triplets, quadruplets, and quintuplets comprised 94.4%, 5.1%, 0.5%, and 0.07% respectively. Complete matching and linkage was achieved for 322,162 (98.8%) of the records. These represent 304,466 twins, 16,068 triplets, 1448 quadruplets, and 180 quintuplets. The distribution of selected sociodemographic characteristics of the mothers is shown in Table 1. Figure 1 displays crude mortality rates for twins, triplets, quadruplets, and quintuplets; the absolute counts of early mortality and its components are shown in Table 2. Table 3 summarizes the adjusted estimates for early mortality and its components with twins used as the referent category. Overall, triplets, quadruplets, and quintuplets had higher risks of mortality than twins. Generally, the twins versus higher-multiple disparity in mortality tended to be lower for stillborn fetuses than for neonatal and infant mortality. The most pronounced discrepancy was for quintuplets, among whom the risk for mortality was 4 –15 times that of twins. In contrast to the modest differences in risk levels observed for triplets and quadruplets, quintuplets demonstrated a wide gap between intrauterine and extrauterine mortality risks (Table 3). Table 4 shows the results for the various mortality indices when triplets were used as the referent category and comparisons made with higher-order births, and Table 4. Mortality Estimates by Plurality Groups, Obtained With the Generalized Estimating Equation Framework With Triplet Category as Referent Group Quadruplets Stillborn fetus Neonatal mortality Infant mortality Early mortality
Quintuplets
AOR*
95% CI
AOR*
95% CI
2.2 1.2 1.2 1.4
1.3, 3.6 0.8, 1.8 0.8, 1.7 1.1, 1.9
2.5 5.3 4.9 4.4
0.6, 10.0 2.2, 12.7 2.1, 11.3 2.0, 9.5
Abbreviations as in Table 3. * Adjusted estimates were generated by taking into account the confounding effects of maternal age, race, education, parity, adequacy of prenatal care, marital status, smoking, and alcohol ingestion during pregnancy.
682
Salihu et al
Early Mortality Among Multiples
Table 5 contains the proportion of excess mortality that could be prevented by replacing higher-order multiple gestations with triplets and twins through the restriction of the number of transferred embryos. The proportion of excess mortality that could be prevented among triplets by replacing them with twins ranged from 45% to 64%, with a lower magnitude of preventive fraction for intrauterine demise and higher values for extrauterine death. Among quadruplets, the level of preventable deaths that could be achieved by replacing them with twins was around 70%, with practically no variation across mortality types. The most remarkable benefit of limitation was clearly in the case of quintuplets. If they were replaced by twins, the avoidable excess mortality would be in the range of 74 –93%. The right side of Table 5 features the proportion of excess mortality preventable by replacing quadruplets and quintuplets with triplets instead of twins. The level of preventable deaths was still higher for quintuplets (60 – 81%) compared with that for quadruplets (15–54%). DISCUSSION This study provides risk estimates for early mortality as well as the magnitude of excess deaths potentially preventable by avoiding higher-order multiple gestation through the restriction of the number of transferred embryos during assisted reproductive technology cycles. We found that the likelihood of stillborn fetus and neonatal and infant mortality increased in a dose-dependent fashion with each additional fetus. In general, there are very limited data on mortality outcomes among higher-order births beyond triplets and quadruplets.20,21 An important limitation in these previous studies is the lack of adjustment for several potentially confounding characteristics, which were considered and controlled for in our study. Thus, it is reasonable to assume that the risk estimates reported in this study provide the most comprehensive and minimally biased risk estimates for stillborn fetus and neonatal and infant mortality among twins, triplets, quadruplets, and quintuplets to date in the
OBSTETRICS & GYNECOLOGY
Table 5. Preventive Fraction for Stillborn Fetuses, Neonatal Mortality, Perinatal Mortality, and Infant Mortality by Plurality Groups Preventive fraction by limiting to triplets
Preventive fraction by limiting to twins Stillborn fetus Neonatal mortality Infant mortality Early mortality
Triplets
Quadruplets
Quintuplets
Quadruplets
Quintuplets
169 (44.8) 570 (64.3) 625 (62.4) 794 (57.5)
37 (70.4) 67 (70.9) 72 (69.1) 109 (69.4)
7 (74.2) 38 (93.4) 39 (92.5) 46 (88.5)
28 (53.7) 15 (16.0) 16 (15.3) 44 (28.0)
6 (60.0) 33 (81.2) 33 (79.5) 39 (75.0)
Data are presented as n (%).
United States. To that extent, the results of our study should be useful for counseling women considering infertility treatment options. An important aspect of this study is the calculated preventive fraction for early mortality. These are adjusted estimates for the numbers of deaths that could be prevented by avoiding the creation of higher-order multiple gestations and replacing them with either twins or triplets. This information should help to inform the current debate regarding the need for averting higherorder multiple gestations by restricting the number of transferred embryos.5 A previous study has used data from the United Kingdom to examine the cost effectiveness of two- versus threeembryo transfers and found the benefits to be comparable.22 A cost– benefit or cost– utility analysis of this nature, especially, for higher multiples in the United States, though clearly beyond the scope of this article, is indicated. Such a study could characterize the costs and benefits of preventing higher-order births through the restriction of the number of transferred embryos, one potential disadvantage of which is a reduced pregnancy success rate. However, the ability to culture embryos to the blastocyst stage23 now permits the transfer of as few as two embryos with pregnancy occurring in up to 60% of women.9,10 Assuming this pregnancy rate to be fixed, then to avoid higher-order multiples, the number of transferred blastocysts could be limited to a maximum of two to three with substantially high pregnancy success rates. Currently, the American Society for Reproductive Medicine recommends the transfer of two good quality embryos for patients with favorable prognosis, three for above-average, four for average, and no more than five for below-average cases.24 In other industrialized countries, the maximum number of transferable embryos is lower than that recommended by the American Society for Reproductive Medicine. The British Fertility Society recommends a maximum of two embryos.25 The European Society for Human Reproduction and Embryology26 advocates either two or three, depending on female age. Other fertility experts have strongly suggested
VOL. 102, NO. 4, OCTOBER 2003
that we should be moving toward the transfer of only one embryo.27 Hence, our recommendation of a maximum of two to three blastocysts based on the analysis above seems reasonable. A clear limitation in this study is our lack of information on assisted reproduction procedures. This information is not available in the birth record of infants born in the United States. As a result, we were unable to conduct separate analyses for artificially and naturally conceived multiple gestations, However, currently almost all quadruplets and quintuplets are the results of assisted reproduction technology,20 and because the focus of our analysis was mainly on these plurality groups it is unlikely that information on whether these pregnancies were natural or artificial would have influenced our findings. On the other hand, triplets and twins in this study were conceived either spontaneously or through assisted reproductive technology. It is therefore likely that the risk estimates obtained with twins and triplets used as referent categories could have been impacted by this heterogeneous composition. Another limitation of this study is the lack of information on those multiples that underwent reduction procedures to lower-order multiples. In other words, it is possible that some of the triplets and twins we used as referents were actually the results of reduction procedures of previous higher-order multiples. Currently, the available information suggests that survivors of reduction procedures face either higher-than-expected risks of morbidity or mortality, or, at best, similar risks as twins and triplets that did not result from reduction procedures.28,29 Therefore, it is likely that our computed values for preventive fractions represent conservative estimates. In summary, our findings, based on comprehensive US data, support that there would be substantial, quantifiable benefit in preventing the occurrence of higherorder multiple gestations and provide a sound rationale for limiting the number of embryos transferred during assisted reproduction procedures.
Salihu et al
Early Mortality Among Multiples
683
REFERENCES 1. Kiely JL, Kiely M. Epidemiologic trends in multiple births in the United States, 1997-1998. Twin Res 2001;14:131–3. 2. Blickstein I, Goldman RD, Mazkereth R. Incidence and birth weight characteristics of twins born to mothers aged 40 years or more compared with 35-39 years old mothers: A population study. J Perinat Med 2001;29:128–32. 3. Ventura SJ. Trends and variations in first births to older women, 1970-86. National Center for Health Statistics. Vital Health Stat 1989;47:1–27. 4. Martin JA, Park MM. Trends in twin and triplet births: 1980-97. Natl Vital Stat Rep 1999;47:1–6. 5. Evans MI, Kramer RL, Yaron Y, Drugan A, Johnson MP. What are the ethical and technical problems associated with multifetal pregnancy reduction? Clin Obstet Gynecol 1998;41:47–54. 6. Torok O, Lapinski R, Salafia CM, Bernasko J, Berkowitz RL. Multifetal pregnancy reduction is not associated with an increased risk of intrauterine growth restriction, except for very-high-order multiples. Am J Obstet Gynecol 1998; 179:221–5. 7. Templeton A, Morris JK. Reducing the risk of multiple births by transfer of two embryos after in vitro fertilization. N Engl J Med 1998;339:573–7. 8. Gleicher N, Oleske D, Tur-Kaspa I, Vidali A, Karande V. Reducing the risk of high order multiple pregnancy after ovarian stimulation with gonadotropins. N Engl J Med 2000;343:2–7. 9. Gardner DK, Vella P, Lane M, Wagley L, Schlenker T, Schoolcraft WB. Culture and transfer of human blastocysts increases implantation rates and reduces the need for multiple embryo transfers. Fertil Steril 1998;69:84–8. 10. Gardner D, Schoolcraft WB, Wagley L, Schlenker T, Stevens J, Hesla J. A prospective randomized trial of blastocyst culture and transfer in in-vitro fertilization. Hum Reprod 1998;13:3434–40. 11. Martin J, Curtin S, Saulnier M, Mousavi J. Development of the matched multiple birth file. In: 1995-1997 matched multiple birth dataset. NCHS CD-ROM series 21, no. 12. Hyattsville, Maryland: National Center for Health Statistics, 2000. 12. National Center for Health Statistics. 1995-1997 linked birth/infant death data set. Vital statistics of the United States: Quality control procedures. Hyattsville, Maryland: US Department of Health and Human Services, Centers for Disease Control and Prevention, 17. 13. Alexander GR, Kotelchuck M. Quantifying the adequacy of prenatal care: A comparison of indices. Public Health Rep 1996;3:408–18. 14. Alexander GR, Cornely DA. Prenatal care utilization: Its measurement and relationship to pregnancy outcome. Am J Prev Med 1987;3:243–53. 15. Kogan MD, Martin JA, Alexander GR, Kotelchuck M, Ventura SJ, Frigoletto FD. The changing pattern of prena-
684
Salihu et al
Early Mortality Among Multiples
16.
17.
18. 19.
20. 21.
22.
23.
24.
25. 26.
27.
28.
29.
tal care utilization in the United States, 1981-1995, using different prenatal care indices. JAMA 1998;279:1623–8. Buescher PA, Taylor KP, Davis MH, Bowling JM. The quality of the new birth certificate data: A validation study in North Carolina. Am J Public Health 1993;83:1163–5. DiGiuseppe DL, Aron DC, Ranbom L, Harper DL, Rosenthal GE. Reliability of birth certificate data: A multihospital comparison to medical records information. Matern Child Health J 2002;16:169–79. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986;42:121–30. Hennekens CH, Buring JE. Measures of disease frequency and association. In: Mayrent SL, ed. Epidemiology in medicine, 1st ed. Boston: Little, Brown and Company, 1987:88. Francois K, Alperin A, Elliott JP. Outcomes of quintuplet pregnancies. J Reprod Med 2001;46:1047–51. Skrablin S, Kuvacic I, Pavicic D, Kalafatic D, Goluza T. Maternal neonatal outcome in quadruplet and quintuplet versus triplet gestations. Eur J Obstet Gynecol 2000;88: 147–52. Keith LG, Blickstein I, Oleszczuk JJ, Keith DM, eds. Triplet pregnancies and their consequences. London: The Parthenon Publishing Group, 2002. Gardner DK, Lane M. Culture of viable human blastocysts in defined sequential serum-free media. Hum Reprod 1998;13:148–59. American Society for Reproductive Medicine. Guidelines on number of embryos transferred: A Practice Committee report: a committee opinion. Birmingham, Alabama: American Society for Reproductive Medicine, 1999. Murdoch AP. How many embryos should be transferred? Hum Reprod 1998;13:2666–70. ESHRE Capri Workshop. Infertility revisited: The state of the art today and tomorrow. Hum Reprod 1996;11: 1779–807. Coetsier T, Dhont M. Avoiding multiple pregnancies in IVF: Who’s afraid of single embryo transfer? Hum Reprod 1998;13:2663–4. Papageorghiou AT, Liao AW, Skentou C, Sebire NJ, Nicolaides KH. Trichorionic triplet pregnancies at 10-14 weeks: Outcome after embryo reduction compared to expectant management. J Matern Fetal Neonatal Med 2002;11: 307–12. Jirsova S, Mardesic T, Muller P, Huttelova R, Zvarova J, Jirkovsky M. Multi-fetal pregnancy reduction does not influence perinatal results in twin pregnancies. Twin Res 2001;4:422–5.
Address reprint requests to: Hamisu Salihu, MD, PhD, University of Alabama at Birmingham, Department of Maternal and Child Health, 1665 University Boulevard, Room 320, Birmingham, Alabama 35294; E-mail: [email protected]. Received April 14, 2003. Received in revised form June 15, 2003. Accepted June 26, 2003.
OBSTETRICS & GYNECOLOGY