- Email: [email protected]
Perinatal mortality in twins and singletons matched for gestational age at delivery at ≥30 weeks
Perinatal mortality in twins and singletons matched for gestational age at delivery at >30 weeks SarahJ. Kilpatrick, MD, PhD," Rebecca Jackson, MD," and Mary S. Croughan-Minihane, PhD b San Francisco, California OBJECTIVE: Our purpose was to compare perinatal mortality between twins and singletons matched for gestational age at delivery with the hypothesis that perinatal mortality would be similar. STUDY DI-'SIGN: A retrospective cohort study compared perinatal mortality in twins (790 babies) matched for gestational age at delivery at >_30weeks wRh the next singleton. Perinatal mortality was also compared by expected growth (small and average for gestational age) and by the cause of the preterm delivery (medical indication, spontaneous preterm delivery, or preterm premature rupture of membranes with preterm delivery). RESULTS: Uncorrected perinatal mortality was significantly greater in singletons (56/1000) than in twins (26/1000) (/3 = 0.004). After major congenital anomalies were excluded, perinatal mortality remained significantly higher in singletons (25/1000) than in twins (11/1000) (p = 0.04). In preterm deliveries perinatal mortality was significantly greater in singletons (72/10013)than in twins (19/1000) (p = 0.0002). Perinatal mortality was significantly greater in small- Compared with average-for-gestational-ageneonates, whether singletons or twins (p = 0.005), and singleton small-for-gestational-age neonates had a significantly greater perinatal mortality (105/1000) than did twin small-for-gestational-age neonates (36/1000) (p = 0.02). Perinatal mortality was not significantly different between preterm singletons and twins delivered by spontaneous preterm labor. However, perinatal mortality in preterm singletons delivered for a medical indication (156/1000) was significantly greater than for twins (20/1000) (p = 0.0006). CONCLUSIONS: At _>30weeks at delivery, twins either had a perinatal mortality similar to, or less than, singletons matched for gestational age at delivery. This suggests that when preterm delivery is controlled for perinatal mortality is not greater in twins. In twins, without risk factors for perinatal death, routine antepartum testing may not be indicated. (AM J OBSTETGYNECOL1996;174:66-71 .)
Key words: Twins, perinatal mortality
It is well known that twins have a significantly higher perinatal mortality than do singletons. Perinatal mortality has been reported as 4 to 10 times greater in twins compared with singletons) :~ Presumably because of this increased perinatal mortality, it has been advocated that routine antepartum testing including nonstress testing (NST) and biophysical profile be used in the management of twins.~7 The rationale given in The American College of Obstetricians and Gynecologists Technical Bulletin on multiple gestations is that testing in twins is "reliable and practical ''s and in a recent obstetrics textbook " . . . feasible and efficacious. ''8' v However, although the purpose of antepartum testing is to decrease perinatal mortality, it may not be logical to assume that all twins need routine antepartum testing. From the Departments of Obstetrics, Gynecology, and Reproductive Sciences" and Family and Community Medicine and Epidemiology and Biostatistics, b University of California, San Francisco. Received for publication January 8, 1995; revised May 25, 1995; acceptedJ'une 5, 1995, Reprint requests: S.J. Kilpatriek, MiD, PhD, Box 0132, University of California, San Francisco, San Francisco, CA 94143. Copyright © 1996 by Mosby-Year Book, Inc. 0002-9378/96 $5.00+ 0 6/1/66800 66
The high perinatal mortality in twins is largely associated with the increased rate of low birth weight deliveries from preterm delivery and delivery of small-for-gestational-age (SGA) babies. I'' " Antepartum testing will not prevent preterm delivery nor prevent the development of intrauterine growth retardation (IUGR). With appropriate interval ultrasonography for fetal growth, the diagnosis of IUGR should be made anteparmm, and at that time the patient should be referred for antepartum testing because IUGR is a known risk factor for perinatal mortality.I2 In the absence of IUGR it is not clear that twin gestations have a higher risk for perinatal mortality compared with singletons of similar gestational age. However, routine antepartum testing has been adopted in twins in spite of the lack of a significant difference in perinatal mortality in tested versus untested twins. 7 In this time of escalating health costs, it is more important than ever to carefully evaluate the use and efficacy of routine procedures. To justify, routine antepartum fetal surveillance, it is necessary to document an increased risk of perinatal death in twins without other risk factors for perinatal mortality such as IUGR, preeclampsia, or diabetes compared with singletons at the same gestational age
Yblume 174, Nmnber 1, Part 1 AmJ Obstet Gynecol
at delivery. Specifically, the purpose of this study was to compare perinatal mortality between twins and singletons matched for gestational age at delivery >29 weeks, a gestational age before which it would be rare to perform antepartum testing. Our hypothesis was that there would be no difference in perinatal mortality between twins and singletons compared in this fashion. We further hypothesized that the cause of preterm birth would affect perinatal mortality and that more singletons than twins would be delivered preterm for medical indications, whereas more twins than singletons would be delivered preterm spontaneously. To test this hypothesis, perinatal mortality was compared by cause of preterm birth. Patients and methods A retrospective cohort study was performed comparing perinatal mortality in twins to that in gestational agematched singletons by use of the University of California San Francisco, perinatal database. Data from all deliveries at >20 weeks' gestation since January 1976 have been entered in the database. Data collected in the database include antepartum, intrapartum, postpartum, and neonatal variables. Intrapartum variables are recorded at delivery and all maternal and neonatal charts are reviewed after maternal and neonatal discharge to abstract remaining data. Gestational age is determined by the best obstetric estimate including last menstrual period, size-dates concordancy, and ultrasonography results if performed at <24 weeks. The obstetric estimate is then compared with the neonatal estimate. If there is a >2-week discrepancy or a discrepancy that straddles 36 to 37 weeks, the chart is reviewed by a perinatologist to obtain the best estimate of dates. The neonatal estimate is used as the best date only if the obstetric estimate does not include either an ultrasonography at <24 weeks or a first prenatal examination equal to dates at <24 weeks. Further, at delivery all preterm births are categorized as spontaneous preterm labor and delivery, premature rupture of membranes with preterm delivery, or medically indicated preterm birth. For the purposes of this study, after triplets reduced to twins ( n = 5) were excluded, all twins at _>30 completed weeks' gestation at delivery from January 1976 through July 1993 ( n = 395 sets, 790 babies) were matched for gestational age at delivery (_+1 week) to the next two singleton deliveries. Gestational age >30 weeks was chosen a priori because it is rare that antepartum testing would begin before 30 weeks. Severe congenital anomalies identified antenatally were then excluded from both twins and singletons because this would represent a group that would likely not require anteparmm testing ( n = 24 babies). Anomalies excluded included trisomies 13, 18, and 21; conjoined twins; congenital diaphragmatic hernia, hydrancephaly; anencephaly, prune-belly syndrome; thanatophoric dwarfism; and Potter's syn-
Kilpatrick, Jackson, and Cmughan-Minihane
67
drome. Perinatal mortality was defined as stillbirth or neonatal death at _<30 days of age per 1000 deliveries because this was the definition used at the inception of the perinatal database in 1976. Uncorrected perinatal mortality and perinatal mortality after exclusion of antenatally identified severe anomalies were compared between twins and singletons. The effects of chorionicity and the presence or absence of growth retardation on perinatal mortality were also determined. SGA was defined as birth weight at <10th percentile, average for gestational age (AGA) as birth weight at the 10th to 90th percentile, and large for gestational age (LGA) as birth weight >90th percentile for sex and gestational age at delivery by California standards. ~3Singleton growth charts were used because that is standard practice. In the twins discordant growth was defined as a 20% difference in birth weight between the twins. The frequency of discordancy was determined. Perinatal mortality was compared between discordant and nondiscordant twins and between discordant twins and singletons. Analysis of variance and Student t test were used for comparing means of continuous data, whereas Z2 statistics were used for determining associations between categoric variables. A p value <0.05 was considered statistically significant. Results From January 1976 to July 1993, there were 28,417 births and 395 sets of twins (790 babies) delivered at >30 weeks of gestation. Placental chorionicity based on operative report and or pathologic examination was 3% monochorionic monoamniotic, 25% monochorionic diamniotic, 62% dichorionic diamniotic, and 10% unknown. The uncorrected perinatal mortality in singletons was significantly higher than that in twins matched for gestational age at delivery (56/1000 vs 26/1000, respectively) (p = 0.004). Table I presents the descriptive data for twins and singletons after antenatally identified severe anomalies were excluded. There were no statistically significant differences between singletons and twins in maternal age, parity, or frequency of maternal transport, but there was a significantly higher proportion of nonwhite women among singletons. Gestational age at delivery was identical in {wins and singletons because it was matched a priori. Not surprisingly, birth weight was significantly less in the twins compared with singletons. In spite of exclusion of antenatally identified major anomalies, 5.4% of singletons and 2.2% of {wins had major anomalies diagnosed after delivery (p = 0.001). Overall perinatal mortality remained significantly greater in singletons than in twins, even after antenatally identified severe anomalies were excluded in both singletons and twins, with a perinatal mortality of 42:1000 and 13:1000 for singletons and twins, respectively (Table II).
68 Kilpatrick, Jackson, and Croughan-Minihane
January 1996 AmJ Obstet Gynecol
Table I. Maternal a n d neonatal demographic a n d descriptive characteristics in singletons and twins after antenatally identified anomalies were excluded
No. Maternal age (yr) (mean -+SD) Nulliparous (%) Ethnicity (%) White Black Asian Hispanic Other Maternal transport (%) Gestational age at delivery (wk) (mean +_SD) Birth weight (gm) (metal -+SD)
Singletons
Twins
[
Significance
766 27.3 + 6.0 50
766 27.9 -+5.8 47
46 14 14 15 11 25 35.8 _+3.0
54 16 10 13 7 23 35.8 + 3.0
p - 0.44 p = 1.0
2661 -+ 770
2357 _+613
p = 0.0005
p - 0.05 p = 0.47 p = 0.903
Table II. Deaths a n d perinatal mortality (per 1000 live births) by gestational age in singletons a n d twins ,Singletons
Twins
Gestational cage
Deaths*
PNM
Deaths
PNM
Significance
>37 wk (n = 338) <37 wk (n = 428) TOTAL
1 31 32
3 72 42
2 8 l0
6 19 13
p = 0.56 p = 0.0002 p = 0.0005
. PNM, Perinatal mortality. *Includes stillbirths and neonatal deaths. This difference in perinatal mortality persisted in preterm neonates and in those categorized as low birth weight (<2500 gm) (Tables II a n d III). In very-low-birthweight (<1500 gin) neonates the perinatal mortality was greater in singletons (206/1000) than in twins (91/1000), b u t did n o t reach statistical significance (p = 0.05). At term perinatal mortality was n o t significantly different between twins (3/1000) a n d singletons (6/1000) (Table II). Stillbirths accounted for 50% of perinatal mortality in both singletons a n d twins. Although the perinatal mortality d o u b l e d in monochorionic (26/1000) compared with dichorionic (13/1000) twins, it was n o t statistically significant (p = 0.36). W h e n all major anomalies (antenatally and postnatally identified) were excluded, the overall perinatal mortality was still significantly greater in singletons (25/1000) compared with twins (11/1000) (p = 0.04). Table 1V displays the distribution of SGA, AGA, a n d LGA neonates in the twins a n d singletons a n d compares perinatal mortality a m o n g groups. As expected, there were significantly more SGA twins than SGA singletons, with 25% and 10%, respectively, being SGA. SGA neonates had a significantly higher perinatal mortality than AGA n e o nates whether a product of a singleton (105/1000 vs 32/1000) or a twin gestation (36/1000 vs 5/1000). Further, singleton SGA neonates had a significantly higher perinatal mortality than did twin SGA neonates. Likewise, singleton AGA neonates had a significantly higher perinatal mortality compared with twin AGA neonates. There were too few LGA twins to make any meaningful comparisons.
Table V depicts the distribution of causes of preterm delivery. As predicted, a significantly greater proportion of the p r e t e r m singletons were delivered for a medical indication (36% of singletons vs 23% of twins). In contrast, significantly more twins were delivered prematurely because of spontaneous p r e t e r m labor (42%) compared with singletons (26%). Perinatal mortality by cause of delivery is shown in Table VI. There was no significant difference in perinatal mortality between singletons and twins delivered with spontaneous p r e t e r m labor. W h e n premature r u p t u r e of m e m b r a n e s or a medical indication was the cause for preterm delivery, singletons had a significantly higher perinatal mortality than twins did. Discordant growth was present in 140 twins (18%), a n d 70% of the perinatal mortality in twins was accounted for by discordant twins. The perinatal mortality in discordant twins (50/1000) was significnatly greater than that in n o n d i s c o r d a n t twins (5/1000) ( p = 0.00002) but was n o t significantly different from singletons (42/1000) (p = 0.66). In spite of exclusion of antenatally diagnosed severe anomalies, 50% and 40% of the singleton and twin neonatal deaths, respectively, were associated with a maj o r anomaly. There were insufficient data available on stillbirths to d e t e r m i n e the cause of death. Comment
W h e n p r e t e r m delivery was controlled for by matching for gestational age at delivery, perinatal mortality in twins was approximately one third that of singletons ( t 3 / I 0 0 0 vs 42/1000). This finding is consistent with the concept that the higher perinatal mortality in twins is related
Kilpatrick, Jackson, and Croughan-Minihane
Volume 174, Number 1, Part 1 AmJ Obstet GynecoI
Table
69
III. Deaths and perinatal mortality (per 1000 live births) by birth weight in singletons and twins Twins
Singletons Birth weight (gin)
Deaths*
<2500
22
I
I
PNM
Deaths
73
8
n = 300 <1500
13
PNM
Significance
19
p = 0,000005
91
p = 0.05
n = 421 206
7
n = 63
n = 77
PNM, Perinatal mortality. *Includes stillbirths and neonatal deaths.
Table IV. Perinatal mortality (per 1000 live births) in singletons and twins by SGA, AGA, and LGA Singletons
No.* No. of deaths PNM'~
Twins
SGA
A GA
LGA
SGA
A GA
LGA
76 8 105
654 21 32
36 3 83
193 7 36
571 3 5
2 0 0
PNM, Perinatal mortality. *p < 0.000005, distribution of SGA, AGA, and LGA between singletons and twins. t0.000005 < p < 0.02, perinatal mortality in singleton SGA versus twin SGA, singleton AGA versus twin AGA, singleton SGA versus singleton AGA, and twin SGA versus twin AGA.
tothe higher incidence of p r e t e r m delivery rather than to an i n h e r e n t p r o b l e m with the twin fetus or neonate. This is not an u n p r e c e d e n t e d finding because similar results have b e e n r e p o r t e d for neonatal death w h e n twins and singletons were c o m p a r e d controlling for birth weight. McCarthy et a]. I4 r e p o r t e d that singletons had a significantly higher, or similar, neonatal death rate than twins at <2500 gm birth weight. Likewise, in a study of >16,000 multiple gestations in New York, Kiely ~1 r e p o r t e d that birth weight-specific neonatal mortality was greater in twins only at very small (501 to 1000 gm) or large (>3000 gm) sizes. At birth weights of 1001 to 3000 gm twins had the same or lower neonatal mortality than singletons. A l t h o u g h intrapartum fetal death was significantly m o r e f r e q u e n t in twins than in singletons w h e n not controlled for birth weight, birth weight-specific intrap a r t u m fetal death at 501 to 2500 g m was significantly less in twins." This agrees with our data in neonates weighing <2500 gm at delivery. Thus, if either gestational age or birth weight are controlled for, twins have less or equal perinatal mortality c o m p a r e d to singletons. Even in SGA twins, a g r o u p e x p e c t e d to have a high perinatal mortality, the perinatal mortality was significantly less than in SGA singletons. This finding m i g h t be explained by the fact that singleton curves were used to diagnose SGA. It has b e e n suggested that an SGA twin may in fact be similar to an AGA singletonJ 5 W h e n the birth weight distributions of twins and singletons are corrected so that the means are e q u a t e d (i.e., converted to a standardized z score), perinatal mortality was h i g h e r in twins. In o t h e r words, these authors argue that a m e a n weight twin (2567 gin) should be c o m p a r e d with the m e a n weight singleton (3364 gm), not to the same size singleton. In contrast to their report, we f o u n d that the
perinatal mortality in AGA singletons was nearly identical to that in SGA twins, at 32:1000 versus 36:1000, respectively. This supports our hypothesis that gestational a g e m a t c h e d singletons and twins do not differ in perinatal mortality. Likewise, it has b e e n suggested that twins mature sooner than singletons, which m i g h t explain the significantly lower perinatal mortality at <37 weeks' gestation in twins. Multiple reports have suggested a h i g h e r perinatal mortality for twins weighing >2500 g m than for singletons and that the incidence of I U G R increases in twins after 38 weeks, suggesting that the twin maturation curve may in fact be shifted to the leftJ ~' 1< ~6 O u r study has several limitations. T h e data may be generalizable only to a similar clinical institution, that is, a tertiary care center that cares for high-risk and p r e t e r m fetuses and neonates. T h e overall perinatal mortality for singletons for our facility, 42:1000, is h i g h e r than that r e p o r t e d in the literature, which varies from 9 to 24: 1000J °' TM ~' ~ This may be explained by the fact that the previously r e p o r t e d perin atal mortality was from large city, state, or national statistics that incorporated m o r e low-risk pregnancies than our data did. Approximately 25% of both our singletons and twins were maternal transports, indicating that either there was a medical p r o b l e m for the m o t h e r or fetus or the pregnancy was p r e t e r m . T h a t the perinatal mortality was largely limited to p r e t e r m neonates is supported by the fact that there was only one death in term singletons, for a perinatal mortality of 3: 1000. Interestingly, the overall perinatal mortality in the twins (13:1000) was lower than what has b e e n r e p o r t e d (14 to 107 : 1000).~-:~ These reports were all based on data from 1968-1986 and included many nontertiary care centers and may reflect p o o r e r antenatal
70 Kilpatrick, Jackson, and Croughan-Minihane
January 1996 AmJ Obstet Gynecol
Table V. P r e t e r m delivery causes for singletons and twins
Cause
Singletons
Twins
Significance
No. Spontaneous (%) PROM (%) Medically indicated (%)
428 27 37 36
428 42 35 23
p = 0.00001 p = 0.44 p = 0.0001
PROM, Premature rupture of membranes.
Table VI. Perinatal mortality (per 1000 live births) by cause of p r e t e r m delivery in singletons and twins
Singletons Cause
Deaths
Spontaneous PROM Medically indicated
2 5 24
][
Twins PNM
Deaths
PNM
Significance
18 31 156
6 0 2
33 0 20
p = 0.43 p = 0.03 p = 0.0006
PNM, Perinatal mortality; PROM, premature rupture of membranes.
recognition of twins and p o o r e r o u t c o m e for p r e t e r m deliveries. The increased perinatal mortality in singletons could be explained by the fact that, as predicted, m o r e singletons than twins were delivered p r e t e r m for a medical indication. In fact, the perinatal mortality was the highest in singletons delivered for a medical indication. This would also r e p r e s e n t a group of patients, whether twin or singleton, who would require a n t e p a r t u m testing because of the medical problem. The most p e r t i n e n t comparison groups to assess whether twins inherently have a h i g h e r perinatal mortality would be those patients who were delivered p r e t e r m with spontaneous labor only. In fact, there was no difference in perinatal mortality between p r e t e r m singletons and twins delivered for spontaneous p r e t e r m labor. W h e n m a t c h e d for gestational age at delivery and whether analyzed by gestational age, birth weight, presence of SGA, or cause of p r e t e r m delivery, perinatal mortality in twins was always equal to or significantly less than in singletons. This suggests that there is no a priori reason to routinely p e r f o r m a n t e p a r t u m testing on twin pregnancies without o t h e r risk factors for perinatal death, Additional evidence that twins without o t h e r risk factors for perinatal mortality do n o t n e e d testing is that there was no significant differences in perinatal mortality in tested c o m p a r e d with untested twins. 7 Further, the majority of babies with a b n o r m a l testing and a perinatal loss had risk factors including IUGR, twin-twin transfusion syndrome, preeclampsia, or diabetes, all of which should have b e e n diagnosed a n t e p a r t u m and should have p r o m p t e d a n t e p a r t u m testing. 4' 5.7. ~v.2oNot surprisingly, a higher perinatal mortality was r e p o r t e d in twins with a nonreactive NST c o m p a r e d with a reactive NST. ~ Howeve~, the two perinatal deaths in their nonreactive group
had antenatally identified risk factors o t h e r than multiple gestation: severe I U G R with chronic maternal alcohol use and twin-twin transfusion syndrome. In addition, 90% of the twins without a n t e p a r t u m complications had only reactive NSTs, suggesting that twins without a n t e p a r t u m complications may n o t n e e d a n t e p a r t u m testing] 9 As expected, the presence of SGA was m o r e c o m m o n and was associated with a significantly increased perinatal mortality in twins so, unequivocally, suspected I U G R should be an indication for a n t e p a r t u m testing. In addition, as r e p o r t e d elsewhere, the presence of discordancy was associated with an increase in perinatal mortality and should be an indication for antenatal testing in twins, s Thus the ability to antenatally diagnose I U G R and discordancy b e c o m e s of utmost importance. Unfortunately, we were n o t able to test how well suspected I U G R correlated with SGA. A r e m a i n i n g c o n c e r n is that either ultrasonography would miss I U G R w h e n present or that p o o r growth would occur so rapidly as to preclude recognition by even timely ultrasonographic examination. Questions for future research include what is the best time interval and how often is it necessary to evaluate twins for growth by ultrasonography and w h e t h e r it is m o r e cost effective to p e r f o r m regular ultrasonographic examinations or routine a n t e p a r t u m testing in twins. O n the basis of these results, it may be reasonable not to p e r f o r m routine antepartum testing in twins unless there is a medical risk factor , IUGR, or discordant growth. T h e cost effectiveness and safety of this suggestion require prospective evaluation. REFERENCES
1. Hawrylyshyn PA, Barkin M, Bernstein A, Papsin FR. Twin pregnancies--a continuing perinatal challenge. Obstet Gynecol 1982;59:463-6. 2. Kovacs BW, Kirschbaum TH, Paul RH. Twin gestations, I:
Volume 174, Number 1, Part 1
Kilpatrick, Jackson, and Croughan-Minihane
71
Am J Obstet Gyneco]
3. 4. 5. 6. 7. 8. 9.
10. 11.
antenatal care and complications. Obstet Gynecol 1989;74: 313-7. Spellacy WN, Handler A, Ferre CD. A case-control study of 1253 twin pregnancies from a 1982-1987 perinatal database. Obstet Gynecol 1990;75:168-71. Bailey D, Flynn AM, Kelly J, O'Conor M. Amtepartum fetal heart rate monitoring in multiple pregnancy. Br J Obstet Gynaecol 1980;87:561-4. Devoe LD, Azor H. Simultaneous nonstress fetal heart rate testing in twin pregnancy. Obstet Gynecol 1981;58:450-5. Lodeiro JG, Vintzileos AM, Feinstein SJ, Campbell WA, Nochimson DJ. Fetal biophysical profile in twin gestations. Obstet Gynecol 1986;67:824-7. Sherman SJ, Kovacs BW, Medearis AL, Bear MB, Paul RH. Nonstvess test assessment of twins. J Reprod Med 1992;37: 804-7. American College of Obstetricans and Gynecologists. Multiple gestation. Washington: American College of Obstetricians and Gynecologists, 1989; Technical bulletin no 131. Chitkara U, Berkowitz RL. Multiple gestations. In: Gabbe SG, Nieby]JR, SimpsonJL, eds. Obstetrics: normal and problem pregnancies. New York: Churchill Livingston, 1991:881921. Ghai V, Vidyasagar D. Morbidity and mortality factors in twins: an epidemiologic approach. Clin Perinatol 1988;15: 123-40. KielyJL. The epidemiology of perinatal mortality in multiple births. Bull N YAcad Med 1990;66:618-37,
12. Gabbe SG. Intrauterine growth retardation. In: Gabbe SG, NiebylJR, SimpsonJL, eds. Obstetrics: normal and problem pregnancies. New York: Churchill Livingstone, 1991:92344. 13. Cunningham GC, Hawes WE, Madore C, Norris FD, Williams RL. Intrauterine growth and neonatal risk in California. Sacramento, California: Infant Health Section, Maternal and Child Health, Department of Health, 1976. 14. McCarthy BJ, Sachs BP, Layde PM, Burton A, Terry JS, Rochat R. The epidemiology of neonatal death in twins. A.,a J OBSTETG'eYECOL1981;141:252-6. 15. Buekens P, Wilcox A. Why do small twins have a lower mortality rate than small singletons? A_,aJ OBSTETG'e,rECOL 1993; 168:937-4l. 16. Luke B, Witter FR, Abbey I--I, et al. Gestational age-specific birth weights of twin versus singletons, Acta Genet Med Gemellol 1991;40:69-76. 17. Clarkson JE, Buckfield PM, Herbison GE The outcome of twin pregnancies in Dunedin 1968-1978. Aust N Z J Obstet Gynaecol 1982;22:127-30. 18. Ho SK, Wu PYK. Perinatal factors and neonatal morbidity in twin pregnancy. A ~ J OBSTETGV~'ECOL1975;122:979-87. 19. Blake GD, Knuppel RA, Ingardia CJ, Lake M, Aumann G, Hanson M. Evaluation ofnonstress fetal heart rate testing in multiple gestations. Obstet Gynecol 1984;63:528-32. 20. Knuppel RA, Rattan PK, ScerboJC, O'Brien ~VF. Intrauterine fetal death in twins after 32 weeks gestation. Obstet Gynecol 1985;65:172-5.
Key words: Twins, perinatal mortality
It is well known that twins have a significantly higher perinatal mortality than do singletons. Perinatal mortality has been reported as 4 to 10 times greater in twins compared with singletons) :~ Presumably because of this increased perinatal mortality, it has been advocated that routine antepartum testing including nonstress testing (NST) and biophysical profile be used in the management of twins.~7 The rationale given in The American College of Obstetricians and Gynecologists Technical Bulletin on multiple gestations is that testing in twins is "reliable and practical ''s and in a recent obstetrics textbook " . . . feasible and efficacious. ''8' v However, although the purpose of antepartum testing is to decrease perinatal mortality, it may not be logical to assume that all twins need routine antepartum testing. From the Departments of Obstetrics, Gynecology, and Reproductive Sciences" and Family and Community Medicine and Epidemiology and Biostatistics, b University of California, San Francisco. Received for publication January 8, 1995; revised May 25, 1995; acceptedJ'une 5, 1995, Reprint requests: S.J. Kilpatriek, MiD, PhD, Box 0132, University of California, San Francisco, San Francisco, CA 94143. Copyright © 1996 by Mosby-Year Book, Inc. 0002-9378/96 $5.00+ 0 6/1/66800 66
The high perinatal mortality in twins is largely associated with the increased rate of low birth weight deliveries from preterm delivery and delivery of small-for-gestational-age (SGA) babies. I'' " Antepartum testing will not prevent preterm delivery nor prevent the development of intrauterine growth retardation (IUGR). With appropriate interval ultrasonography for fetal growth, the diagnosis of IUGR should be made anteparmm, and at that time the patient should be referred for antepartum testing because IUGR is a known risk factor for perinatal mortality.I2 In the absence of IUGR it is not clear that twin gestations have a higher risk for perinatal mortality compared with singletons of similar gestational age. However, routine antepartum testing has been adopted in twins in spite of the lack of a significant difference in perinatal mortality in tested versus untested twins. 7 In this time of escalating health costs, it is more important than ever to carefully evaluate the use and efficacy of routine procedures. To justify, routine antepartum fetal surveillance, it is necessary to document an increased risk of perinatal death in twins without other risk factors for perinatal mortality such as IUGR, preeclampsia, or diabetes compared with singletons at the same gestational age
Yblume 174, Nmnber 1, Part 1 AmJ Obstet Gynecol
at delivery. Specifically, the purpose of this study was to compare perinatal mortality between twins and singletons matched for gestational age at delivery >29 weeks, a gestational age before which it would be rare to perform antepartum testing. Our hypothesis was that there would be no difference in perinatal mortality between twins and singletons compared in this fashion. We further hypothesized that the cause of preterm birth would affect perinatal mortality and that more singletons than twins would be delivered preterm for medical indications, whereas more twins than singletons would be delivered preterm spontaneously. To test this hypothesis, perinatal mortality was compared by cause of preterm birth. Patients and methods A retrospective cohort study was performed comparing perinatal mortality in twins to that in gestational agematched singletons by use of the University of California San Francisco, perinatal database. Data from all deliveries at >20 weeks' gestation since January 1976 have been entered in the database. Data collected in the database include antepartum, intrapartum, postpartum, and neonatal variables. Intrapartum variables are recorded at delivery and all maternal and neonatal charts are reviewed after maternal and neonatal discharge to abstract remaining data. Gestational age is determined by the best obstetric estimate including last menstrual period, size-dates concordancy, and ultrasonography results if performed at <24 weeks. The obstetric estimate is then compared with the neonatal estimate. If there is a >2-week discrepancy or a discrepancy that straddles 36 to 37 weeks, the chart is reviewed by a perinatologist to obtain the best estimate of dates. The neonatal estimate is used as the best date only if the obstetric estimate does not include either an ultrasonography at <24 weeks or a first prenatal examination equal to dates at <24 weeks. Further, at delivery all preterm births are categorized as spontaneous preterm labor and delivery, premature rupture of membranes with preterm delivery, or medically indicated preterm birth. For the purposes of this study, after triplets reduced to twins ( n = 5) were excluded, all twins at _>30 completed weeks' gestation at delivery from January 1976 through July 1993 ( n = 395 sets, 790 babies) were matched for gestational age at delivery (_+1 week) to the next two singleton deliveries. Gestational age >30 weeks was chosen a priori because it is rare that antepartum testing would begin before 30 weeks. Severe congenital anomalies identified antenatally were then excluded from both twins and singletons because this would represent a group that would likely not require anteparmm testing ( n = 24 babies). Anomalies excluded included trisomies 13, 18, and 21; conjoined twins; congenital diaphragmatic hernia, hydrancephaly; anencephaly, prune-belly syndrome; thanatophoric dwarfism; and Potter's syn-
Kilpatrick, Jackson, and Cmughan-Minihane
67
drome. Perinatal mortality was defined as stillbirth or neonatal death at _<30 days of age per 1000 deliveries because this was the definition used at the inception of the perinatal database in 1976. Uncorrected perinatal mortality and perinatal mortality after exclusion of antenatally identified severe anomalies were compared between twins and singletons. The effects of chorionicity and the presence or absence of growth retardation on perinatal mortality were also determined. SGA was defined as birth weight at <10th percentile, average for gestational age (AGA) as birth weight at the 10th to 90th percentile, and large for gestational age (LGA) as birth weight >90th percentile for sex and gestational age at delivery by California standards. ~3Singleton growth charts were used because that is standard practice. In the twins discordant growth was defined as a 20% difference in birth weight between the twins. The frequency of discordancy was determined. Perinatal mortality was compared between discordant and nondiscordant twins and between discordant twins and singletons. Analysis of variance and Student t test were used for comparing means of continuous data, whereas Z2 statistics were used for determining associations between categoric variables. A p value <0.05 was considered statistically significant. Results From January 1976 to July 1993, there were 28,417 births and 395 sets of twins (790 babies) delivered at >30 weeks of gestation. Placental chorionicity based on operative report and or pathologic examination was 3% monochorionic monoamniotic, 25% monochorionic diamniotic, 62% dichorionic diamniotic, and 10% unknown. The uncorrected perinatal mortality in singletons was significantly higher than that in twins matched for gestational age at delivery (56/1000 vs 26/1000, respectively) (p = 0.004). Table I presents the descriptive data for twins and singletons after antenatally identified severe anomalies were excluded. There were no statistically significant differences between singletons and twins in maternal age, parity, or frequency of maternal transport, but there was a significantly higher proportion of nonwhite women among singletons. Gestational age at delivery was identical in {wins and singletons because it was matched a priori. Not surprisingly, birth weight was significantly less in the twins compared with singletons. In spite of exclusion of antenatally identified major anomalies, 5.4% of singletons and 2.2% of {wins had major anomalies diagnosed after delivery (p = 0.001). Overall perinatal mortality remained significantly greater in singletons than in twins, even after antenatally identified severe anomalies were excluded in both singletons and twins, with a perinatal mortality of 42:1000 and 13:1000 for singletons and twins, respectively (Table II).
68 Kilpatrick, Jackson, and Croughan-Minihane
January 1996 AmJ Obstet Gynecol
Table I. Maternal a n d neonatal demographic a n d descriptive characteristics in singletons and twins after antenatally identified anomalies were excluded
No. Maternal age (yr) (mean -+SD) Nulliparous (%) Ethnicity (%) White Black Asian Hispanic Other Maternal transport (%) Gestational age at delivery (wk) (mean +_SD) Birth weight (gm) (metal -+SD)
Singletons
Twins
[
Significance
766 27.3 + 6.0 50
766 27.9 -+5.8 47
46 14 14 15 11 25 35.8 _+3.0
54 16 10 13 7 23 35.8 + 3.0
p - 0.44 p = 1.0
2661 -+ 770
2357 _+613
p = 0.0005
p - 0.05 p = 0.47 p = 0.903
Table II. Deaths a n d perinatal mortality (per 1000 live births) by gestational age in singletons a n d twins ,Singletons
Twins
Gestational cage
Deaths*
PNM
Deaths
PNM
Significance
>37 wk (n = 338) <37 wk (n = 428) TOTAL
1 31 32
3 72 42
2 8 l0
6 19 13
p = 0.56 p = 0.0002 p = 0.0005
. PNM, Perinatal mortality. *Includes stillbirths and neonatal deaths. This difference in perinatal mortality persisted in preterm neonates and in those categorized as low birth weight (<2500 gm) (Tables II a n d III). In very-low-birthweight (<1500 gin) neonates the perinatal mortality was greater in singletons (206/1000) than in twins (91/1000), b u t did n o t reach statistical significance (p = 0.05). At term perinatal mortality was n o t significantly different between twins (3/1000) a n d singletons (6/1000) (Table II). Stillbirths accounted for 50% of perinatal mortality in both singletons a n d twins. Although the perinatal mortality d o u b l e d in monochorionic (26/1000) compared with dichorionic (13/1000) twins, it was n o t statistically significant (p = 0.36). W h e n all major anomalies (antenatally and postnatally identified) were excluded, the overall perinatal mortality was still significantly greater in singletons (25/1000) compared with twins (11/1000) (p = 0.04). Table 1V displays the distribution of SGA, AGA, a n d LGA neonates in the twins a n d singletons a n d compares perinatal mortality a m o n g groups. As expected, there were significantly more SGA twins than SGA singletons, with 25% and 10%, respectively, being SGA. SGA neonates had a significantly higher perinatal mortality than AGA n e o nates whether a product of a singleton (105/1000 vs 32/1000) or a twin gestation (36/1000 vs 5/1000). Further, singleton SGA neonates had a significantly higher perinatal mortality than did twin SGA neonates. Likewise, singleton AGA neonates had a significantly higher perinatal mortality compared with twin AGA neonates. There were too few LGA twins to make any meaningful comparisons.
Table V depicts the distribution of causes of preterm delivery. As predicted, a significantly greater proportion of the p r e t e r m singletons were delivered for a medical indication (36% of singletons vs 23% of twins). In contrast, significantly more twins were delivered prematurely because of spontaneous p r e t e r m labor (42%) compared with singletons (26%). Perinatal mortality by cause of delivery is shown in Table VI. There was no significant difference in perinatal mortality between singletons and twins delivered with spontaneous p r e t e r m labor. W h e n premature r u p t u r e of m e m b r a n e s or a medical indication was the cause for preterm delivery, singletons had a significantly higher perinatal mortality than twins did. Discordant growth was present in 140 twins (18%), a n d 70% of the perinatal mortality in twins was accounted for by discordant twins. The perinatal mortality in discordant twins (50/1000) was significnatly greater than that in n o n d i s c o r d a n t twins (5/1000) ( p = 0.00002) but was n o t significantly different from singletons (42/1000) (p = 0.66). In spite of exclusion of antenatally diagnosed severe anomalies, 50% and 40% of the singleton and twin neonatal deaths, respectively, were associated with a maj o r anomaly. There were insufficient data available on stillbirths to d e t e r m i n e the cause of death. Comment
W h e n p r e t e r m delivery was controlled for by matching for gestational age at delivery, perinatal mortality in twins was approximately one third that of singletons ( t 3 / I 0 0 0 vs 42/1000). This finding is consistent with the concept that the higher perinatal mortality in twins is related
Kilpatrick, Jackson, and Croughan-Minihane
Volume 174, Number 1, Part 1 AmJ Obstet GynecoI
Table
69
III. Deaths and perinatal mortality (per 1000 live births) by birth weight in singletons and twins Twins
Singletons Birth weight (gin)
Deaths*
<2500
22
I
I
PNM
Deaths
73
8
n = 300 <1500
13
PNM
Significance
19
p = 0,000005
91
p = 0.05
n = 421 206
7
n = 63
n = 77
PNM, Perinatal mortality. *Includes stillbirths and neonatal deaths.
Table IV. Perinatal mortality (per 1000 live births) in singletons and twins by SGA, AGA, and LGA Singletons
No.* No. of deaths PNM'~
Twins
SGA
A GA
LGA
SGA
A GA
LGA
76 8 105
654 21 32
36 3 83
193 7 36
571 3 5
2 0 0
PNM, Perinatal mortality. *p < 0.000005, distribution of SGA, AGA, and LGA between singletons and twins. t0.000005 < p < 0.02, perinatal mortality in singleton SGA versus twin SGA, singleton AGA versus twin AGA, singleton SGA versus singleton AGA, and twin SGA versus twin AGA.
tothe higher incidence of p r e t e r m delivery rather than to an i n h e r e n t p r o b l e m with the twin fetus or neonate. This is not an u n p r e c e d e n t e d finding because similar results have b e e n r e p o r t e d for neonatal death w h e n twins and singletons were c o m p a r e d controlling for birth weight. McCarthy et a]. I4 r e p o r t e d that singletons had a significantly higher, or similar, neonatal death rate than twins at <2500 gm birth weight. Likewise, in a study of >16,000 multiple gestations in New York, Kiely ~1 r e p o r t e d that birth weight-specific neonatal mortality was greater in twins only at very small (501 to 1000 gm) or large (>3000 gm) sizes. At birth weights of 1001 to 3000 gm twins had the same or lower neonatal mortality than singletons. A l t h o u g h intrapartum fetal death was significantly m o r e f r e q u e n t in twins than in singletons w h e n not controlled for birth weight, birth weight-specific intrap a r t u m fetal death at 501 to 2500 g m was significantly less in twins." This agrees with our data in neonates weighing <2500 gm at delivery. Thus, if either gestational age or birth weight are controlled for, twins have less or equal perinatal mortality c o m p a r e d to singletons. Even in SGA twins, a g r o u p e x p e c t e d to have a high perinatal mortality, the perinatal mortality was significantly less than in SGA singletons. This finding m i g h t be explained by the fact that singleton curves were used to diagnose SGA. It has b e e n suggested that an SGA twin may in fact be similar to an AGA singletonJ 5 W h e n the birth weight distributions of twins and singletons are corrected so that the means are e q u a t e d (i.e., converted to a standardized z score), perinatal mortality was h i g h e r in twins. In o t h e r words, these authors argue that a m e a n weight twin (2567 gin) should be c o m p a r e d with the m e a n weight singleton (3364 gm), not to the same size singleton. In contrast to their report, we f o u n d that the
perinatal mortality in AGA singletons was nearly identical to that in SGA twins, at 32:1000 versus 36:1000, respectively. This supports our hypothesis that gestational a g e m a t c h e d singletons and twins do not differ in perinatal mortality. Likewise, it has b e e n suggested that twins mature sooner than singletons, which m i g h t explain the significantly lower perinatal mortality at <37 weeks' gestation in twins. Multiple reports have suggested a h i g h e r perinatal mortality for twins weighing >2500 g m than for singletons and that the incidence of I U G R increases in twins after 38 weeks, suggesting that the twin maturation curve may in fact be shifted to the leftJ ~' 1< ~6 O u r study has several limitations. T h e data may be generalizable only to a similar clinical institution, that is, a tertiary care center that cares for high-risk and p r e t e r m fetuses and neonates. T h e overall perinatal mortality for singletons for our facility, 42:1000, is h i g h e r than that r e p o r t e d in the literature, which varies from 9 to 24: 1000J °' TM ~' ~ This may be explained by the fact that the previously r e p o r t e d perin atal mortality was from large city, state, or national statistics that incorporated m o r e low-risk pregnancies than our data did. Approximately 25% of both our singletons and twins were maternal transports, indicating that either there was a medical p r o b l e m for the m o t h e r or fetus or the pregnancy was p r e t e r m . T h a t the perinatal mortality was largely limited to p r e t e r m neonates is supported by the fact that there was only one death in term singletons, for a perinatal mortality of 3: 1000. Interestingly, the overall perinatal mortality in the twins (13:1000) was lower than what has b e e n r e p o r t e d (14 to 107 : 1000).~-:~ These reports were all based on data from 1968-1986 and included many nontertiary care centers and may reflect p o o r e r antenatal
70 Kilpatrick, Jackson, and Croughan-Minihane
January 1996 AmJ Obstet Gynecol
Table V. P r e t e r m delivery causes for singletons and twins
Cause
Singletons
Twins
Significance
No. Spontaneous (%) PROM (%) Medically indicated (%)
428 27 37 36
428 42 35 23
p = 0.00001 p = 0.44 p = 0.0001
PROM, Premature rupture of membranes.
Table VI. Perinatal mortality (per 1000 live births) by cause of p r e t e r m delivery in singletons and twins
Singletons Cause
Deaths
Spontaneous PROM Medically indicated
2 5 24
][
Twins PNM
Deaths
PNM
Significance
18 31 156
6 0 2
33 0 20
p = 0.43 p = 0.03 p = 0.0006
PNM, Perinatal mortality; PROM, premature rupture of membranes.
recognition of twins and p o o r e r o u t c o m e for p r e t e r m deliveries. The increased perinatal mortality in singletons could be explained by the fact that, as predicted, m o r e singletons than twins were delivered p r e t e r m for a medical indication. In fact, the perinatal mortality was the highest in singletons delivered for a medical indication. This would also r e p r e s e n t a group of patients, whether twin or singleton, who would require a n t e p a r t u m testing because of the medical problem. The most p e r t i n e n t comparison groups to assess whether twins inherently have a h i g h e r perinatal mortality would be those patients who were delivered p r e t e r m with spontaneous labor only. In fact, there was no difference in perinatal mortality between p r e t e r m singletons and twins delivered for spontaneous p r e t e r m labor. W h e n m a t c h e d for gestational age at delivery and whether analyzed by gestational age, birth weight, presence of SGA, or cause of p r e t e r m delivery, perinatal mortality in twins was always equal to or significantly less than in singletons. This suggests that there is no a priori reason to routinely p e r f o r m a n t e p a r t u m testing on twin pregnancies without o t h e r risk factors for perinatal death, Additional evidence that twins without o t h e r risk factors for perinatal mortality do n o t n e e d testing is that there was no significant differences in perinatal mortality in tested c o m p a r e d with untested twins. 7 Further, the majority of babies with a b n o r m a l testing and a perinatal loss had risk factors including IUGR, twin-twin transfusion syndrome, preeclampsia, or diabetes, all of which should have b e e n diagnosed a n t e p a r t u m and should have p r o m p t e d a n t e p a r t u m testing. 4' 5.7. ~v.2oNot surprisingly, a higher perinatal mortality was r e p o r t e d in twins with a nonreactive NST c o m p a r e d with a reactive NST. ~ Howeve~, the two perinatal deaths in their nonreactive group
had antenatally identified risk factors o t h e r than multiple gestation: severe I U G R with chronic maternal alcohol use and twin-twin transfusion syndrome. In addition, 90% of the twins without a n t e p a r t u m complications had only reactive NSTs, suggesting that twins without a n t e p a r t u m complications may n o t n e e d a n t e p a r t u m testing] 9 As expected, the presence of SGA was m o r e c o m m o n and was associated with a significantly increased perinatal mortality in twins so, unequivocally, suspected I U G R should be an indication for a n t e p a r t u m testing. In addition, as r e p o r t e d elsewhere, the presence of discordancy was associated with an increase in perinatal mortality and should be an indication for antenatal testing in twins, s Thus the ability to antenatally diagnose I U G R and discordancy b e c o m e s of utmost importance. Unfortunately, we were n o t able to test how well suspected I U G R correlated with SGA. A r e m a i n i n g c o n c e r n is that either ultrasonography would miss I U G R w h e n present or that p o o r growth would occur so rapidly as to preclude recognition by even timely ultrasonographic examination. Questions for future research include what is the best time interval and how often is it necessary to evaluate twins for growth by ultrasonography and w h e t h e r it is m o r e cost effective to p e r f o r m regular ultrasonographic examinations or routine a n t e p a r t u m testing in twins. O n the basis of these results, it may be reasonable not to p e r f o r m routine antepartum testing in twins unless there is a medical risk factor , IUGR, or discordant growth. T h e cost effectiveness and safety of this suggestion require prospective evaluation. REFERENCES
1. Hawrylyshyn PA, Barkin M, Bernstein A, Papsin FR. Twin pregnancies--a continuing perinatal challenge. Obstet Gynecol 1982;59:463-6. 2. Kovacs BW, Kirschbaum TH, Paul RH. Twin gestations, I:
Volume 174, Number 1, Part 1
Kilpatrick, Jackson, and Croughan-Minihane
71
Am J Obstet Gyneco]
3. 4. 5. 6. 7. 8. 9.
10. 11.
antenatal care and complications. Obstet Gynecol 1989;74: 313-7. Spellacy WN, Handler A, Ferre CD. A case-control study of 1253 twin pregnancies from a 1982-1987 perinatal database. Obstet Gynecol 1990;75:168-71. Bailey D, Flynn AM, Kelly J, O'Conor M. Amtepartum fetal heart rate monitoring in multiple pregnancy. Br J Obstet Gynaecol 1980;87:561-4. Devoe LD, Azor H. Simultaneous nonstress fetal heart rate testing in twin pregnancy. Obstet Gynecol 1981;58:450-5. Lodeiro JG, Vintzileos AM, Feinstein SJ, Campbell WA, Nochimson DJ. Fetal biophysical profile in twin gestations. Obstet Gynecol 1986;67:824-7. Sherman SJ, Kovacs BW, Medearis AL, Bear MB, Paul RH. Nonstvess test assessment of twins. J Reprod Med 1992;37: 804-7. American College of Obstetricans and Gynecologists. Multiple gestation. Washington: American College of Obstetricians and Gynecologists, 1989; Technical bulletin no 131. Chitkara U, Berkowitz RL. Multiple gestations. In: Gabbe SG, Nieby]JR, SimpsonJL, eds. Obstetrics: normal and problem pregnancies. New York: Churchill Livingston, 1991:881921. Ghai V, Vidyasagar D. Morbidity and mortality factors in twins: an epidemiologic approach. Clin Perinatol 1988;15: 123-40. KielyJL. The epidemiology of perinatal mortality in multiple births. Bull N YAcad Med 1990;66:618-37,
12. Gabbe SG. Intrauterine growth retardation. In: Gabbe SG, NiebylJR, SimpsonJL, eds. Obstetrics: normal and problem pregnancies. New York: Churchill Livingstone, 1991:92344. 13. Cunningham GC, Hawes WE, Madore C, Norris FD, Williams RL. Intrauterine growth and neonatal risk in California. Sacramento, California: Infant Health Section, Maternal and Child Health, Department of Health, 1976. 14. McCarthy BJ, Sachs BP, Layde PM, Burton A, Terry JS, Rochat R. The epidemiology of neonatal death in twins. A.,a J OBSTETG'eYECOL1981;141:252-6. 15. Buekens P, Wilcox A. Why do small twins have a lower mortality rate than small singletons? A_,aJ OBSTETG'e,rECOL 1993; 168:937-4l. 16. Luke B, Witter FR, Abbey I--I, et al. Gestational age-specific birth weights of twin versus singletons, Acta Genet Med Gemellol 1991;40:69-76. 17. Clarkson JE, Buckfield PM, Herbison GE The outcome of twin pregnancies in Dunedin 1968-1978. Aust N Z J Obstet Gynaecol 1982;22:127-30. 18. Ho SK, Wu PYK. Perinatal factors and neonatal morbidity in twin pregnancy. A ~ J OBSTETGV~'ECOL1975;122:979-87. 19. Blake GD, Knuppel RA, Ingardia CJ, Lake M, Aumann G, Hanson M. Evaluation ofnonstress fetal heart rate testing in multiple gestations. Obstet Gynecol 1984;63:528-32. 20. Knuppel RA, Rattan PK, ScerboJC, O'Brien ~VF. Intrauterine fetal death in twins after 32 weeks gestation. Obstet Gynecol 1985;65:172-5.