Outcome of Very Low Birth Weight Infants: Multiple Gestation Versus Singletons

(1)

Outcome

of Very

Low

Multiple

Gestation

Birth

Weight

Infants:

Versus

Singletons

Carol H. Leonard, PhD*; Robert E. Piecuch, MD*; Roberta A. Ballard, MD*; and Bruce A. B. Cooper, PhD

ABSTRACT. Objective. Multiple gestation infants are

overrepresented in intensive care nurseries, and have been reported to have greater morbidity than singletons. A cohort of very low birth weight infants was examined to determine outcome of premature infants based on ges-tation type (multiple or single) and hypothesized that at

this low birth weight, the outcome of the groups would be

similar.

Method. The sample was composed of all infants with

birth weights 125O g born in a 10-yearperiod (September

1977 through September 1987). Ninety-two percent (n = 364) of the infants discharged were seen at 1 year of age, and 73% (n = 249) were observed to school age.

Mor-bidity was assessed by neurodevelopmental

examina-lions and standard developmental tests.

Results. At 1 year of age and at school age, there were no differences in neurologic or neurosensory outcome

be-tween multiple gestation and single gestation infants.

Lo-gistic regression analyses were performed on the school

age data, using cognitive outcome as the dependent van-able and gestation type, birth weight, gestational age, in-tracranial hemorrhage, chronic lung disease, and a social

risk factor as predictor variables. Gestation type was not associated with cognitive outcome at school age. Social

risk factors and chronic lung disease showed an associa-Hon with cognitive outcome at school age.

Conclusions. Multiple gestation was not related to

in-creased morbidity in this very low birth weight group.

The developmental outcome of all infants with birth

weights 1250 g in this study was related to medical and

social risk factors. These findings were consistent for a

large group of infants over a 10-year period. Pediatrics

1994;93:611-615; multiple gestation, very low birth weight, development, outcome.

ABBREVIATIONS. MGI, multiple gestation infant; SES, socioeco-nomic status; ICN, intensive care nursery; SD, standard deviation; ICH, intracranial hemorrhage; CT, computed tomography; CLD, chronic lung disease; SOCR, social risk; CPS, Child Protective Ser-vices; McGCI McCarthy General Cognitive Index; WISC-R FSIQ, Wechsler Intelligence Scale for Children, Revised Full Scale IQ.

Historically, multiple gestation infants (MGI) have

been considered to be at increased risk for mortality

and morbidity in the neonatal period and for

subse-quent delayed growth and development.3 MGI have

From the *Depaj.fuflent of Pediatrics, University of California, San Francisco, CA; the Department of Pediatrics, Children’s Hospital of Philadelphia, Philadelphia, PA; and the §California School of Professional Psychology, Alameda, CA.

Received for publication Apr 15, 1993; accepted Aug 31, 1993.

Reprint requests to (C.H.L.) Dept of Pediatrics, Nursery Follow-Up Pro-gram, Box 0748, University of California, San Francisco, San Francisco, CA

94143.

been studied primarily as twin pairs. Although MGIs

represent I to 2% of all births, they are

overrepre-sented in tertiary level nurseries because of the

in-creased incidence of obstetrical problems, premature

births, and congenital anomalies in multiple gestation

pregnancies. In addition to medical and

developmen-tal risks, families of twins have been identified as

hay-ing a higher rate of abuse/neglect.

During the past 25 years, studies have produced

conflicting findings about morbidity in multiple

ges-tation infants compared with singletons. Some

stud-ies have reported significant neurologic deficits, such

as cerebral palsy, in twins whereas others have

re-ported milder morbidity such as early language delay

and slightly lower IQ scores in twins than in

singletons.3’7 On the other hand, studies that have

carefully controlled for birth weight and/or

gesta-tional age in comparing twins with singletons

gen-erally find no differences in either neonatal morbidity

or subsequent neurologic and developmental

outcome.125 Recently one controlled study reported

increased periventricular leukomalacia in twins.16 In

addition to conflicting findings, some well-controlled

studies have had small samples, or only report

short-term follow-up (<24 months of age).

A few investigators have examined socioeconomic

status (SES) and social risk factors in their studies of

twin populations. Wilson’s longitudinal study of

twins found prematurity to operate as a short-term

suppressor effect on the cognitive development of

twins <1750 g.17 At 6 years of age, twins from upper

SES families had overcome the initial effects of

pre-maturity to perform in a normal or average range on

standardized tests.

Twins

from lower SES families

had lower IQ scores at age 6.

In the following study, medical and social risk

fac-tors were examined in relation to gestation type (MGI

versus singleton) in a group of infants weighing

1250 g at birth, born during a 10-year period

(Sep-tember 1, 1977 through September 1, 1987). It was

hy-pothesized that gestation type would not show

asso-ciation with neurologic, neurosensory, or cognitive

outcome when birth weight was restricted to very low

birth weight. At very low birth weight it was

hypoth-esized that medical and/or social risk factors were

more likely to be associated with poor outcome than

gestation type. The rate of major handicap at 1 year

and at school age for the MGI and singleton groups

was determined. Gestation type, medical risk factors,

and social risk factors were examined for their

rela-tionship to cognitive/developmental outcome at

preschool/school age. Within the multiple gestation

at Viet Nam:AAP Sponsored on September 1, 2020 www.aappublications.org/news

(2)

group of infants, birth order, and survivorship were

examined for relation to long-term outcome.

Sample

METHODS

Four hundred thirteen infants weighing I 250 g were dis-charged from the Intensive Care Nursery (ICN) at Mount Zion Hospital between September 1, 1977 and September 1, 1987. This group was racially diverse: 52% white, 26% African-American, 13% Hispanic, and 8% other, primarily Asian-American. The families of the children in this study came from a range of SES levels but almost half the families (42%) were of low SES, as de-fined by criteria noted below. The infants were part of a longitu-dinal study of outcome of very low birth weight and informed consent for the study was obtained from parents at discharge or the first clinic visit.

Of the 413 infants, 82 were multiple gestation infants, and 329 were singletons. Two infants were excluded from the study for congenital anomalies (one MGI, one singleton). Because infants were selected by birth weight, members of a multiple gestation set

were included in the study only if their birth weight was 1250 g

Table I describes the characteristics of the multiple gestation in-fants. The MGI group was composed primarily of twin deliveries with a small number of triplet deliveries.

The MGI and singleton groups were similar in birth weight, gender distribution, percentage of infants appropriate for gesta-tional age, and percentage of infants admitted after neonatal trans-port (Table 2). Gestational age was determined by Dubowitz ex-amination at birth. The MGI group had a mean gestational age of 28.8weeks (standard deviation (SD) = 2.0) compared with a mean gestational age of 28.3 weeks (SD = 2.0) for the singleton group,

and a lesser degree of intracranial hemorrhage (ICH).

Seventeen infants who were discharged from the ICN to home died during the study. Fifteen infants died during the first year of life. Acquired immunodeficiency syndrome was diagnosed in one infant in the first year who ultimately died, and one infant had an accidental death at age 3.

At I year, 30 surviving infants had been lost to follow-up, leaving 364 infants for study (92%). For the school-age sample, 62 infants were lost, and 41 had not yet reached school age. Twelve infants were not included in the school-age logistic regression analysis because they did not have an early computed tomogra-phy (CT) scan or ultrasound examination. Therefore, 73% (249

infants) of the original surviving group were available for the

school age analysis. The characteristics of the infants in the study at 1 year and at school age did not differ from the characteristics of the infants at discharge.

Risk Factors

Medical and social risk factors were determined for each infant enrolled in the study.

Chronic Lung Disease (CLD). Infants were given a score for CLD

reflecting the number of days they required 02. The scores were determined as: 0 = <30 days of 02; 1= 30 to 60 days of 02, 2 = 61 to 90 days of 02; and 3 = >90 days of 02. Infants with a score of

2 or 3 (>60 days of 02) were designated as having significant CLD. Sixty days of 02 was chosen as a marker for CLD rather than

TABLE 1. Characteristics of Multiple Gestation Infants

Mode of delivery, No.

Vaginal vertex 45

Vaginal breech 13

Cesarean section 24

Birth weights in multiple gestation sets, %

Concordant 82

Discordant 13

. Unreported 5

Birth order, No.

Infant A (first born) 38

Infant B (second born) 42

Infant C (third born) 2

Survivorship in multiple gestation sets, No.

Single survivor in set 26

Survivor of complete set 56

30 days, because of the very low birth weight of this sample. With mean birth weight at approximately 1000 g and gestational age of 28 weeks, the median 02 requirement for the whole group was 30 days.

ICH. Infants had CT scans (1977 through 1980) or ultrasonog-raphy (1981 through 1987) within the first week after delivery, with studies repeated weekly if there were abnormalities. ICH were graded in the nursery according to the four-level grading system described by Papile et al)8 For statistical analyses, grades I and II were considered as one group and grades III and IV were considered as a second group. One infant with bilateral periven-tricular leukomalacia was placed with the group containing grades III and IV ICH for the purposes of this study.

SES. Families were considered to be of low SES if one or more of the following were present: maternal education <12 years, health insurance obtained from public assistance, or primary in-come earner in family was unemployed. Forty-two percent of families in the study were of low SES by this rating.

Social Risk (SOCR). Referral to the Department of Child

Protec-tive Services (CPS) of the Department of Social Services for abuse and/or neglect with resultant action (removal from home, remain in home with CPS supervision, remain in home with additional social services obtained) was considered a social risk factor. Re-ferrals were primarily for mild abuse or neglect; however, there were two cases of more significant abuse.

Outcome Measures

Infants were assessed according to a protocol for very low birth weight infants in the Follow-Up Clinic. Infants were seen fre-quently in the first year of life for neurodevelopmental examina-tions. They were assessed by a neonatologist with training in developmental and behavioral pediatrics. Formal developmental testing was conducted by a psychologist at adjusted age 2 months, 18 months, 2#{189}years, and chronologic age 4#{189}years and 7 years. School age outcome was considered the last formal devel-opmental test completed by the child (4#{189}year examination or 7 year examination). Neurologic, neurosensory, and cognitive out-come were considered as follows:

Neurologic Outcome. Cerebral palsy or significant abnormalities in muscle strength, tone, reflexes, or movement patterns causing ftmctional impairment were classified as “abnormal” outcome. These included such diagnoses as hemiplegia, spastic diplegia, quadriplegia, and ataxia. Infants with abnormalities in tone, strength, or reflexes who did not have a clear diagnosis were given a suspect classification. Mild neuromotor abnormalities such as tremors or clumsiness were not included as abnormal outcome.

Neurosensory Outcome. Bilateral blindness or hearing loss re-quiring hearing aids or other communication aids were consid-ered as abnormal neurosensory outcome.

Cognitive Outcome. The McCarthy Scales of Children’s Abilities was administered at age 4#{189};and the Wechsler Intelligence Scale for Children, Revised was administered at age 7. Children were assessed with these instruments if they had the functional ability to complete the assessment. Children with hemiplegias, for ex-ample, were able to routinely participate in assessments. Eight children with severe motor limitations and severely limited

corn-munication abilities were not administered the cognitive assess-ment but were placed in a priori abnormal category for cognitive outcome. All follow-up studies must deal with severely disabled infants and decide how to include or exclude them for long-term follow-up. It has been decided to include them in this abnormal outcome category although they were generally not fully exam-med on cognitive scales after preschool because of their untestable status on standard tests. Reports of functional skills were obtained from their special education programs where possible.

Scores on the McCarthy General Cognitive Index (MCGCI) and the Wechsler Intelligence Scale for Children, Revised Full Scale IQ (WISC-R FSIQ) that were between one and two standard devia-tions below the mean were considered mildly abnormal; scores more than two standard deviations below the mean were consid-ered moderately to severely abnormal. Children with severe mo-tor impairment were placed in this category because of their significant special educational requirements.

(3)

corn-TABLE 2. Sample Characteristics of Multiple Gestation Infants and Singletons at Discharge

Multiple Gestation Infant Singleton P

No. 82 329

Birth weight, g 1015 ± 162 998 ± 165 NS

Gestational age, wk 28.85 ± 2.07 28.33 ± 2.05+ < .05

SGA*, n (%) 8 (10) 26 (8) NS

Inborn 38 (51) 147 (52) NS

Male 39 (52) 135 (48) NS

Chronic lung disease

02 > 60 d 13 (15.9) 77 (23.4) NS

No intracranial hemorrhage 63 (81) 212 (664 < .05

Low SES 29 (39) 134 (48) NS

*SGA, small for gestational age; SES, socioeconomic status; NS, not significant; df, degrees of freedom. t t = 2.08, 409 df p = .04.

t6.92,2dfP= .04.

pleted both assessments, not all children did. Correlations were obtained for children who did complete both assessments. The correlation coefficient between the MCCCI and the WISC-R FSIQ was .90 for the MGI sample, and .80 for the singleton sample. Where both assessments had been obtained, WISC-R FSIQ was used in analyses.

Statistical Analyses

Bivariate analyses (pooled variance t tests or x tests) were performed on the two groups for the variables of birth weight, gestational age, inborn/outborn status, gender, and appropriate weight for gestational age variables. These analyses of the MGI and singleton groups were performed for three time periods: dis-charge, 1-year follow-up, and school-age follow-up to assure that the groups remained representative of the original sample and that obtained differences were not due to sample attrition.

x2analyses were used to examine differences in neurologic and neurosensory outcome at I and 5 years. For the school age sample, logistic regression analyses were performed to determine the best predictors of school age cognitive outcome. Outcome within the multiple gestation group was examined by x analyses for differ-ences related to birth order and survivorship.

Because this hypothesis predicted no differences in develop-mental outcome between MGI and singleton groups, appropriate sample size to detect a difference, if one existed, was of concern. The sample size for analyses of sample characteristics at three time periods (discharge, I year, 5 years) and neurologic outcome at I and 5 years (t tests, analyses) was large enough to yield a power level >80. A standard power analysis for logistic regression has not yet been developed. At this point statisticians estimate what the power would be based on analogous statistical methods, for example, if instead of a logistic regression model, a multiple re-gression model with five predictors was used to examine cognitive

outcome at 5 years, the power of the analysis would be 1 .0. A categorical analysis of the 5-year data (2 x 2 tables) would yield a power level of .96.

RESULTS

Discharge

Medical and social risk factors at discharge are

pre-sented in Table 2. There was no significant difference

in the incidence of CLD (02 >60 days). There was a

significant difference in the presence of ICH between

groups. The MGI group had significantly less ICH

than the singleton group. This difference,

accumu-lated over a 10-year period, was not explainable by

significantly different rates of ICH in infants who died

in the ICN. A subanalysis of the rate of ICH in infants

who died before discharge showed that ICH occurred

at similar rates in both multiple and single gestation

infants who lived longer than 1 day but who did not

survive to discharge. Infants surviving <24 hours

were generally critically ill or extremely immature

and not likely to have received scans or ultrasounds.

Of the infants who survived >24 hours, and received

scans or ultrasounds before expiration, 7/10 (70%) of

MGI and 55/72 (76%) of singletons had some grade

of ICH.

The MGI group had a slight but significant

differ-ence in gestational age (MGI = 28.85 weeks

gesta-tional age, singletons - 28.33 weeks gestational age)

at similar mean birth weights. The gestational age

variable was entered into a logistic regression to

de-termine if this small difference had an effect on

out-come. Further analyses of the distribution of birth

weight and gestational age for AGA MGI and

single-tons and SGA MGI and singletons were conducted

using the Kolmogorov-Smirnov 2 sample test for

equality of distributions. There were no differences in

the distributions of the variables birth weight and

ges-tational age in the MGI and singleton AGA and SGA

groups. This indicates that at similar birth weights,

MGI infants were judged by the Dubowitz

examina-tion to be slightly older than singleton infants. This

may reflect placental factors that led to accelerated

maturation of the multiple gestation fetus, and/or

smaller weight gain for age in the multiple gestation

fetus.

One-Year Outcome

One-year outcome data were examined for

signifi-cant neurologic or neurosensory handicap.

Neurologic Outcome. Table 3 presents the

neuro-logic outcome at I year adjusted age. There were no

significant differences between MGI and singleton

groups for neurologic deficit. In the MGI group, 5%

had a diagnosis of cerebral palsy, compared with 6%

in the singleton group. Three percent in the MGI

group and 2% in the singleton group had a suspect

neurologic examination.

Neurosensory Outcome. One percent of infants in

each group experienced significant neurosensory

deficits (Table 3). These infants had bilateral blindness

or severe hearing loss.

School Age Outcome

School age cognitive assessments and

neurodevel-opmental assessments were performed at 4#{189}years of

age or 7 years of age. The later examination was used

where available. The mean age of the sample was 77

months (SD, 18.7 months).

Neurologic Outcome. At school age, the incidence of

abnormal neurologic outcome was similar for the

(4)

TABLE 3. Neurologic and Neurosensory Morbidity at One Year and Five Years

One Year Five Years

MGI Singleton

MGI* Singleton

No.

Cerebralpalsy Suspect neurologic

Examination

Blind/deaf

75 4(5%)

2(3%)

I (1%)

289 17(6%)

6(2%)

4 (1%)

52 4(8%)

0(0%)

1 (2%)

197 17(8.6%)

2 (1%)

4 (2%)

*MGI, multiple gestation infant.

MGI and singleton groups. In the MGI group, 8% of

the infants had cerebral palsy or other functional

mo-tor deficit. In the singleton group, 8.6% of the children

had abnormal examinations (Table 3). There were no

new abnormal neurologic cases found in the MGI

group after the 1-year examination. Two singleton

children with abnormal neurologic outcome at 1

year were lost at school age; but two children with

normal examinations at I year of age, had late

pres-entations of spastic diplegia. Two singleton children

had increased tone or reflexes at school age, but did

not have a diagnosis of cerebral palsy, and were

quite functional.

Neurosensory Outcome. The significant

neurosen-sory deficits seen at I year of age were static and

ob-viously remained the same at school age. No new

cases of severe neurosensory deficit were discovered.

Cognitive Outcome. A logistic regression analysis

was performed using cognitive outcome at the 4#{189}or

7 year examination, as the dependent variable. The

later examination was used where available. Five

variables were selected for simultaneous regression.

The predictors entered were: birth weight, ICH, CLD,

gestation type and SOCR. The predictor birth weight

had no unique predictive power and was removed

from the regression. The predictor gestation type was

not significant but it was left in for subsequent

analysis as it was the parameter of interest in this

study. Subsequent analysis using the remaining

pre-dictors showed strong unique relationships for CLD

(P < .002) and SOCR (P < .001). There were weaker

unique associations noted for ICH (P = .127). There

was no association found for the predictor gestation

type (P = .539) (Table 4). A second logistic

regres-sion analysis was performed with the same outcome

variable and predictor variables, but gestational age

was substituted for birth weight. As with birth

weight, gestational age had no unique predictive

power and was removed from the regression.

Outcome Within the Multiple Gestation Infants Group.

In the school age sample of multiple gestation infants,

49% were first born and 51 % were later born.

analy-TABLE 4. Cognitive Outcome at School Age Related to Medi-cal Risk Factors, Social Risk Factor, and Gestation Type

Parameter Standard P Value

Estimate Error

Gestation type -0.20 0.33 .539 NS

ICH* -0.27 0.18 .127t

CLD -0.48 0.13 .002

50CR -1.79 0.42 .001

* ICH, intracranial hemorrhage; CLD, chronic lung disease; 50CR,

social risk.

t In logistic regression models, a probability level of .25 is gener-ally used as a criterion for retaining the variable in the model.’9

ses of the outcome at this age related to birth order

showed no differences between groups (first born

versus later born 2 = 0.03, 1 df, P = .86).

At school age, 64% of the MGI were multiple

sur-vivors and 36% were single survivors (other

mem-bers of the multiple gestation set having expired).

The deaths ranged from the neonatal period to

within the first year of life. There were no

differ-ences in cognitive outcome at school age between

the single survivors and the multiple survivors

(single survivor versus multiple survivor = 0.08,

I df, P = .98).

DISCUSSION

In this study, infants were followed up until school

age with a modest rate (27%) of attrition. The size and

scope of the study reduce the chance fluctuations

pos-sible in short-term follow-up studies with small

sample size. Infants were selected on the basis of very

low birth weight. By restricting the range of birth

weights to 1250 g, it was found that the variable

gestation type was not related to neurologic or

cog-nitive outcome at school age. Although gestation type

was not related to outcome, other risk factors were.

The medical risk factor CLD was related to

long-term cognitive outcome. The criteria of at least 60

days of supplemental 02 appears to have been a

meaningful marker of infants who were more

chronically ill. Some infants in this group were

dis-charged home on 02 and required 02 throughout

the first year of life. Other considerations for

long-term oxygen-dependent infants, such as nutritional

needs, and their effect on cognitive development

were not addressed in this study.

ICH did not show a highly significant relationship

to long-term cognitive outcome, however it fell within

a predetermined range (0 through 25%) of probability

for retention in the logistic regression model.19

Al-though ICH is related to presence of neurologic

defi-cit, in some cases this is primarily a motor dysfunction

that leaves cognitive functions relatively intact. One

child with a hemiplegia, for example, had an IQ in the

superior range at school age.

Children in this study, whether of multiple

gesta-tion or single gestation, had poorer outcome at school

age if they had a CPS referral, a finding consistent

with previous research.2#{176} MGI and singleton infants

had similar rates of referral, which was increased over

that in the general population. The enrollment period

of this study ended coincidentally with the

begin-nings of the epidemic of crack cocaine and multiple

substance abuse. SOCR factors will continue to be

very important in the consideration of factors

(5)

ARTICLES 615

This study found no differences in cognitive

out-come at school age between first born and later born

multiple gestation infants. This constituted a

cross-sectional analysis comparing birth order, not matched

twin pairs, or triplet groups. Comparing

develop-ment of the individuals within pairs or triads would

necessitate a different study, as some infants in sets

were excluded by birth weight >1250 g in this study.

Additionally at this low birth weight, there were

many single survivors of sets, the other set members

having died.

Although a question was posed whether single

sur-vivors of multiple gestation groups might fare better

than multiple survivors, this was not found to be the

case. There were no significant differences in outcome

at school age between single survivors and multiple

survivors. There are many clinical factors that are

in-volved in this question. It has been found, for

ex-ample, that there is wide variation in how families

treat dead infants, depending not only on age of death

(stillborn versus sudden infant death syndrome at 8

months, for example) but also on family dynamics in

the treatment of the memory of the dead infants. Some

families have adaptively integrated the memory of

the dead infant into the fabric of ongoing family life

although other families appear to have a pathological

focus on the dead infant at the expense of surviving

infant(s).

This study addressed outcome of very low birth

weight multiple gestation infants born in a time

period (September 1, 1977 through September 1, 1987)

where multiple births were likely to be naturally

oc-curring, or to be related to maternal drugs to enhance

fertilization. It would not be appropriate to generalize

the results of this study to more recent multiple

ges-tation births where other factors such as in vitro

fertilization with multiple ovum implantation and

possibly fetal reduction surgery may influence the

occurrence or outcome of multiple births.

ACKNOWLEDGMENTS

Support for the statistical analyses in this study was provided by a grant from Perinatal Associates, Inc.

REFERENCES

1. Wenstrom KD, Gall SA. Incidence, morbidity and mortality, and diag-nosis of twin gestations. Clin Perinato!. 1988;15:1-11

2. McCulloch K. Neonatal problems in twins. Cli,, Perinatol. 1988;15:

141-158

3. McDiarmid J Mck, Silva PA. Three-year.old twins and singletons: a comparison of some perinatal, environmental, experiential, and devel-opmental characteristics. Aust Paediatr I.1979;15:243-247

4. Robarge JP, Reynolds ZB, Groothius JR. Increased child abuse in fami-lies with twins. Res Nurs Health. 1982;5:199-203

5. Groothius JR. Altemeier WA, Robarge JP, et al. Increased child abuse in families with twins. Pediatrics. 1982;70:769-773

6. Nelson UB, Martin CA. Increased child abuse in twins. Child Abuse Negl.

1985,9:501-505

7. Gnffiths M. Cerebral palsy in multiple pregnancy. Dev Med Child Neurol. 19679:713-731

8. Hoffman EL, Bennett FC. Birth weight less than 800 grams: changing outcomes and influences of gender and gestation number. Pediatrics.

1990;86:27-34

9. Silva PA, McGee RO, Powell J. Growth and development of twins

compared with singletons at ages five and seven a follow-up report from the Dunedin multidisciplinary child development study. Aust

Paediatr J.1982;18:35-36

10. Silva PA, Crosado B. The growth and development of twins compared with singletons at ages 9 and 11. Aust Paediatr I.1985;21:265-267

11. Yu VYH, Loke HL, Bajuk B, et al. Prognosis for infants born at 23 to 28 weeks’ gestation. Br Med J.1986;293:1200-1203

12. Clarkson JE, Buckfield PM, Herbison GP. The outcome of twin preg-nancies in Dunedin 1968-1978. Aust N Z I Obstret Gynaecol. 1982;22:

127-130

13. Record RG, McKeown T, Edwards JH. An investigation of the difference in measured intelligence between twins and single births. Ann Hum

Genet. 197034:11-20

14. Ghai V, Vidyasagar D. Morbidity and mortality factors in twins-an epidemiologic approach. Cli,z Perinatol. 1988;15:123-140

15. Morley R, Cole TJ, Powell R, Lucas A. Growth and development in premature twins. Arch Dis Child. 1989;64:1042-1045

16. Stafford JR. Collins M,James LS. White matter lesions (WML) in infants of twin gestations. Soc Ped Res Abstr. 1991;29:364A

17. Wilson RS. The Louisville twin study: Developmental synchronies in behavior. Child Dev. 1983;54:298-316

18. Papile L, Burstein J, Burstein R, Koffler H. Incidence and evolution of subependymal and intraventricular hemorrhage: a study of infants with

birth weights less than 1,500 gm. JPediatr. 1978,92:529-534

19. Hosmer DW, Lemershow S. Applied Logistic Regression. New York: Wiley; 1989:82-134

20. Leonard CH, Clyman RI, Piecuch RE, et at. Effect of medical and social risk factors on outcome of prematurity and very low birth weight. IPediatr. 1990;116:620-626

COMPREHENSIVE

COHORT

FOLLOW-UP

IS IMPORTANT

. . . an analysis of the external validity of a trial [generalizability of results] is only

possible if all eligible patients are followed regardless of their randomization

status.

Olschewski M, et at. Analysis of randomized and nonrandomized patients in clinical trials using comprehensive cohort follow-up study design. Controlled C/in Trials. 1992;13:226-239.

Submitted by Student

(6)

1994;93;611

Pediatrics

Carol H. Leonard, Robert E. Piecuch, Roberta A. Ballard and Bruce A. B. Cooper