assessment of in utero hypoxia and risk of sudden infant death syndrome
TRANSCRIPT
Pacdiatric and Perinatal Epidemiology 1989, 3, 157-173
Assessment of in utero hypoxia and risk of sudden infant death syndrome
Germaine M. Buck*, Diane L. Cookfairt, Arthur M. Michalekt, Philip C. NascaS, Susan J. Standfast$ and Lowell E. Sever9 +State University of New York at Buffalo, tRoswell Park Memorial Institute, Buffalo, $New York State Health Department and 5Centerk for Diseuse Control, Atlanta, USA
Summary. Few data are available on the role of hypoxia in sudden infant death syndrome (SIDS). The purpose of this study was to assess whether 10 antenatal factors consistent with in utero hypoxia were associated with an increased risk of SIDS. Cases and two sets of controls were chosen from the Upstate New York Live Birth Cohort for 1974 (n-132948). One hundred and forty-eight SIDS cases were identified, along with 114 dead controls made up of all other sudden deaths. Randomly selected live controls were frequency-matched to cases on mother’s age, race, residence, parity, and infant’s birthdate (n-355). Data were collected from vital certificates (97% response), hospital delivery records (89%), and autopsy reports (100%). Odds ratios and 95% confidence intervals were calculated using Mantel-Haenszel tech- niques and logistic regression. Abnormal uterine bleeding was the only statistically significant (PCO.05) risk factor observed when dead controls were used (OR-5.4). When live controls were used, statistically significant increases in risk were found for: placenta praevia (OR-21.8), abruptio placentae (OR=3.7), multiple birth (OR-29.6), pregnancy interval 112 months (OR-3.8), sexually transmitted disease (OR=6.4), and eclampsia (OR-17.7). These results lend support to a possible hypoxic aetiology of SIDS; however, differences by control p u p suggest that some factors are not specific to SIDS alone but may be risk factors for infant mortality in general.
Address for correspondence: Dr Germaine M. Buck, Department of Social and Preventive Medicine, State University of New York at Buffalo, 2211 Main Street, Building A, Buffalo, New York 14214, USA.
157
158 G.M. Buck et al.
Introduction
Sudden infant death syndrome, commonly referred to as SIDS, is the leading cause of death during postneonatal life.' In the United States, as in most developed countries, the general incidence of SIDS ranges between 1/1000 to 3/1000 live births. Between five and ten thousand infants succumb to SIDS each year in the United States.'-'
SIDS was first recognised as a distinct clinical entity in 1963 and since that time numerous theories have been proposed.' Although the aetiology of SIDS remains unknown, there is some evidence that hypoxia may be associated with SIDS.6
Epidemiologic evidence which in fact supports a hypoxic origin for SIDS includes the following risk factors: maternal smoking during pregnancy,' illicit maternal drug intrauterine growth retardation,IO*I1 and multiple births.I2 Additional support for a hypoxic origin comes from post-mortem studies, and most notably the work of Dr Richard Naeye. Naeye reported seven tissue markers in SIDS cases that were consistent with chronic hyp0xia.l' However, to date, only three of Naeye's tissue markers have been replicated by other investigators: increased periadrenal brown fat," increased hepatic haematopoiesis'' and astroglioisis of the brain.16
The purposes of this study were: (1) to critically assess whether 10 select antenatal variables capable of adversely affecting the fetal environment were significant risk factors for SIDS, and (2) to determine if the results varied according to control group used for comparison. The 10 antenatal variables of interest to this study were: multiple birth; S 1 2 month pregnancy interval; pre-eclampsia; eclampsia; uterine bleeding; vaginitis; sexually transmitted disease; placenta praevia; abruptio placentae; and amnionitis/chorioamnionitis.
The study reported herein is a continuation and elaboration of an earlier descriptive study of thii live birth cohort"*" which suggested that adverse pregnancy-related events may be assodated with SIDS. Specifically, abnornal uterine bleeding during pregnancy, multiple birth, placenta praevia and abruptio placentae were found to be associated with SIDS. These early findings were suggestive of the need for further analytic study to assess the relationship of perinatal hypoxia and risk of SIDS wlde controlling for a variety of potential confounders. Thus, it was the expressed intent of this study to use multivariate techniques to control for confounders in order to assess the relationship between 10 antenatal variables and the subsequent risk of SIDS.
Material and methods
Ascertainment of cases and control A nested case-control study design was used for empirical testing of the 10 antenatal variables under study. The referent population for this study was the
In utero hypoxia and SZDS 159
1974 live birth cohort for Upstate New York resident mothers (exclusive of New York City, n=132948). Computer tapes of birth and death certificates were matched to identify all infants dying between 7 and 365 days after birth and those surviving to their first birthday.
Six hundred and twenty-four infant deaths were identified, of which 321 hospital infant deaths were immediately excluded due to the chronic or severe nature of the cause(s) of death. The remaining 303 deaths occurred suddenly with most infants dying at home, en route to the hospital or shortly upon arrival.
Cause of death as determined by post-mortem examination and as listed on death certificates was initially utilised to categorise these 303 sudden deaths as SIDS or other sudden deaths. SIDS cases were defined according to the Second International Conference on the Causes of Sudden Death in 111fants.l~ Dead controls were all other sudden deaths for whom an explainable cause of death was established after autopsy examination.
One hundred and sixty-one SIDS cases and 142 dead controls were initially identified from the cohort. Thirteen cases and 28 dead controls were excluded due to lack of autopsy examination or incomplete vital certificates. Dead controls represented a heterogeneous group of sudden deaths attributed to the following causes: respiratory infections/complications (n = 62); complications of various diseases (n-24); cardiac irregularities/amhythmias (n = 15); infections (n = 11); and homicides (n=2). Thus, 148 autopsy-confirmed SIDS cases and 114 dead controls were available for study.
A second control group comprised living controls who were randomly selected from the cohort. Live controls were frequency-matched to SIDS cases on infant’s date of birth and mother’s age, race, parity, and county of residence (n=355).
Data collection A standardised data collection instrument was developed specifically for the collection of data from three sources: birth and death certificates, hospital delivery records of mothers and infants, and autopsy reports. All data listed on the vital records were validated using the medical records and autopsy reports as standards, when these documents were available.
Six hundred and seventeen medical record sets (delivery records of mothers and infants) were requested from 139 hospitals throughout New York State. One hundred and thirty-five hospitals participated in the study (97% response), providing 551 medical records sets (89’30 response). Although originally intended, external constraints on the study did not permit the collection of autopsy reports for all sudden infant deaths. As an alternative, 80 autopsy requests were mailed to 31 different counties within Upstate New York when the cause of death-on the death certificate was inconclusive or questionable. All requested autopsy reports were obtained (100’30 response). Three infant deaths listed as SIDS on death certificates were reclassified as dead controls after obtaining autopsy reports; one
160 G.M. Buck et al. Table 1. Sociodemographic characteristics of respondents and non-respondents
Respondents Non-respondents Total Sododemographic characteristics n % n % n %
I
Mother's age tyr) 5 1 8
19-22 23-26 27-30 31-34
1 3 5 Total
Mother's race White Non-white Total
Mother's parity 0 1 2 3+ Total Mother's residence
(persons per square mile) <1-193 194-1238
1239-3541 3542-6389 6390-16925 Total
56 179 144 90 62 20
55 1
436 112 548
137 181 115 118 55 1
116 121 80
113 119 549
10.3 9 32.5 21 26.2 20 16.2 8 11.3 6 3.6 2
100.0 66
79.6 34 20.4 32
100.0 66
24.9 18 32.8 23 20.9 13 21.4 12
100.0 66
21.1 6 22.0 11 14.6 12 20.6 19 21.7 18
100.0 66
13.6 65 31.7 200 30.3 164 12.1 98 9.0 68 3.0 22
100.0 617
51.5 470 48.4 144
100.0 614
27.3 155 34.8 204 19.7 128 18.2 130
100.0 617
9.1 122 16.7 132 18.2 92 28.8 132 27.3 137
100.0 615
10.5 32.4 26.6 15.9 11.0 3.6
100.0
76.5 23.5
100.0
25.1 33.1 20.7 21.1
100.0
19.8 21.5 15.0 21.5 22.3
100.0
dead control was reclassified as a SIDS case. Thus, approximately 5 % of the sample (n-4) was found to be misclassified on the basis of death certificates alone. All four infant deaths were properly categorised for study purposes.
Operational definitions
The 10 antenatal variables under study were hypothesised to be events associated with in utero hypoxia which may act as an insult to the brain stem, although not necessarily via the same aetiologic pathway. As reported in the literature, placenta praevia, abruptio placentae, pre-eclampsia and eclampsia are antenatal events which may threaten placental functioning.20 Antenatal events which challenge the
In utero hypoxia and SZDS 161
development of a good uterine environment possibly via impaired fetoplacental function include: abnormal uterine bleeding during pregnancy, multiple birth and a short interpregnancy interval ( 5 1 2 months).2Q-22 The latter factor has been regarded by many obstetricians as a condition which negatively affects growth and development of the fetus23 and increases neonatal mortality risk.24 Considera- tion of infectious antenatal events (vaginitis, sexually transmitted disease (STD) and amnionitis/chorioamnionitis) was considered important since recent findings suggest that intrauterine infections or septicaemia may result in in utero hypoxia/ ischemh6 Furthermore, it has been speculated that localised infections such as vaginitis or sexually transmitted disease may develop into occult intrauterine in fec t ion~ .~~J~ Further study of these factors appeared warranted, especially given the generalised reduction in immune function and response during pregnancy.
An a 1 y s is
Crude odds ratios and 9546 confidence intervals were calculated using Mantel Haenszel techniques2’ for the 10 antenatal variables under study. Identical but separate analyses were performed for each control group. The potential confounding effects of maternal age, race, parity, and residence were controlled for in the matching process (for live controls only). Additional potential confounders for certain antenatal variables (listed in Tables 5 and 6 ) were controlled for using unconditional maximum likelihood logistic regression techniques described by Breslow & Day.28 All full models for each of the antenatal variables were tested using a backward elimination approach.29 Missing data for individual variables account for some differences in table totals.
Results
Sociodemographic factors
Table I reflects no major differences between respondents or subjects for whom medical records were obtained and non-respondents or subjects for whom no medical records were obtained with the exception of mother’s race and residence. Non-respondents were more likely to be non-white than were respondents (48.4% and 20.4%, respectively). Approximately 43% of respondents and 26% of non-respondents lived in countries with 1238 persons or less per square mile. However, response did not vary by infant status since medical records were obtained for approximately 90% of cases and controls.
Table 2 demonstrates the reliability of the matching process since SIDS cases and live controls were similar with respect to mother’s race, age, parity and county of residence. SIDS cases and unmatched dead controls also were found to be similar with respect to mother’s race, age and parity but differed slightly with
162 G.M. Buck et al. Table 2. Sociodemographic characteristics of sample, by infant status
SIDS Dead controls Live controls Sociodemographic characteristics n Yo n '30 n %
Hospital delivery records Obtained Not obtained Total
Mother's race White Non-white Total
Mother's age <18
18-23 24-29 30-35
>35 Total
Mother's parity 1 2 3 4+ Total
County of mother's residence (persons per square mile) <1-193 194-1238 1239-3541 3542-6389 6390-16925 Total
Mother's education <High school H. S. graduate Attended college Total
130 18
148
115 32
147
6 67 44 27
4 148
32 49 36 31
148
33 32 23 29 31
148
56 62 27
145
87.8 12.2
100.0
78.2 21.8
100.0
4.1 45.3 29.7 18.2 2.7
100.0
21.6 33.1 24.3 20.9
100.0
22.3 21.6 15.5 19.6 20.9
100.0
38.6 42.8 18.6
100.0
100 87.7 14 12.3
114 100.0
90 79.6 23 20.4
113 100.0
11 9.6 48 42.1 35 30.7 17 14.9 3 2.6
114 100.0
35 30.7 31 27.2 26 22.8 22 19.3
114 100.0
17 14.9 20 17.5 15 13.2 31 27.2 31 27.2
114 100.0
47 41.6 44 38.9 22 19.5
113 100.0
321 90.4 34 9.6
355 100.0
265' 74.9 89 25.1
354 100.0
20 5.6 160 45.1 114 32.1 53 14.9
8 2.3 355 100.0
88 24.8 124 34.9
66 18.6 77 21.7
355 100.0
72 20.4 80 22.7 54 15.3 72 20.4 75 21.2
353 100.0
95 27.7 172 50.1 76 22.2
343 100.0
respect to density of mothers' county of residence. Overall, mothers of dead controls were more likely to reside in more populated counties than mothers of SIDS cases. In terms of mothers' educational attainment, mothers of live controls
In utero hypaxia and SZDS 163
had higher educational attainments than mothers of either SIDS cases or dead controls.
A significant difference between SIDS cases and dead controls was observed for age-at-death. The majority of SIDS deaths occurred between 2 and 4 months of age. Dead controls had a more gradual decline in mortality throughout the first year of life. The Mantel trend test revealed a statistically significant (P(O.01) difference with respect to age-at-death for SIDS cases compared to dead controls.
Tables 3 and 4 reflect the crude risk of SIDS related to the 10 antenatal variables under study using dead and live controls, respectively. The original intent of this study was to assess in utero drug and cigarette smoke exposure in addition to the 10 factors. However, the poor recording of maternal smoking and drug usage data in delivery records and the complete absence of these data on birth certificates precluded meaningful analysis of these two variables. However, data were systematically recorded in medical records for the remaining 10 variables and often on birth certificates as well.
When dead controls were used as the comparison group, uterine bleeding during pregnancy was the only Statistically significant risk factor observed and resulted in approximately a 54-fold increase in risk (OR=5.44,95% =1.39,21.35). It should be noted that abnormal uterine bleeding was bleeding not associated with placenta praevia or abruptio placentae. Suggestive but not statistically significant risks were observed in descending order for placenta praevia (OR=6.84), eclampsia (OR=5.52), abruptio placentae (OR=4.58), sexually transmitted disease (OR=3.96), amnionitis/chorioamnionitis (OR = 2.22), mul- tiple birth (OR=2.22), and 5 1 2 month pregnancy interval (OR=2.10). Pre- eclampsia (OR= 0.77) and vaginitis (OR-0.64) during pregnancy were not observed to be elevated risk factors for SIDS in this data set.
When live controls were used as the comparison group, six statistically significant risk factors were observed in the univariate analysis: multiple birth (OR=29.58), placenta praevia (OR=21.80), eclampsia (OR= 17.65), sexually transmitted disease (OR=6.38), 5 1 2 months pregnancy interval (OR=3.77), and abruptio placentae (OR-3.65). Suggestive but not statistically significant risks were observed for amnionitis/chorioamnionitis (OR- 2.36), vaginitis (OR= 1.68), and uterine bleeding during pregnancy (OR= 1.44). History of pre-eclampsia was not observed to be associated with an increased risk of SIDS (OR-0.82) in this data set.
Tables 5 and 6 present the best models obtained through logistic regression for eight of the 10 antenatal variables that had sufficient numbers for study. All full models included gestation. In addition, pre-eclampsia, eclampsia, abruptio placentae, and placenta praevia were adjusted for pregnancy interval. 1 1 2 month pregnancy interval was adjusted for uterine bleeding. Uterine bleeding was adjusted for pregnancy interval and birthweight. Lastly, vaginitis and sexually transmitted disease were adjusted for month prenatal care was initiated.
Tab
le 3
. Risk o
f SI
DS
rela
ted
to a
nten
atal
var
iabl
es, u
sing
dea
d co
ntro
ls fo
r co
mpa
riso
n
cs iz: iY M
ultip
le b
irth
2. p
SIDS
Dea
d co
ntro
ls
Ant
enat
al
Odd
s 95%
conf
iden
ce
vari
able
n
%
n %
ra
tio
inte
rval
w
a-
Yes
11
8.5
4 4.0
, 2.22
(0.70, 7.02)
No
119
91.5
96
96.0
Tot
al
130
100.0
100
100.0
sltm
onth
pre
gnan
cy in
tern
al
Yes
12
11.4
4 5.8
2.10
No
93
88.6
65
94.2
Tot
al
105
100.0
69
100.0
Pre-
ecla
m ps
ia
YeS
N
o T
otal
Ecla
mps
ia
Yes
N
o T
otal
Ute
rine
ble
edin
g Y
eS
No
Tot
al
1 0.8
1 1.0
0.77
129
99.2
99
99.0
130
100.0
100
100.0
3 2.3
-
5.52
-
127
97.7
100
100.0
130
100.0
100
100.0
13
10.0
2 2.0
5.44
117
90.0
98
98.0
130
100.0
100
100.0
(0.66, 6.67)
(0.05, 12.40)
(0.39, 78.61)
(1.39, 21.35)*
Vagi
nitis
Y
es
No
Tot
al
Sexu
ally
tran
smitt
ed d
isea
se
Yes
N
o T
otal
Plac
enta
pra
evia
Y
es
No
Tota
l
Abru
ptio
pla
cent
ae
Yes
N
o T
otal
6 4.6
7 7.0
0.64
124
95.4
93
93.0
130
100.0
100
100.0
5 3.8
1 1 .o
125
96.2
99
99.0
130
100.0
100
100.0
3.96
(0.21, 1.97)
(0.53, 29.67)
4 3.4
-
-
6.84
(0.54, 87.39)
112
96.6
85
100.0
116
100.0
85
100.0
M
6 5.2
1 1.2
4.58
(0.65, 32.50)
7
110
94.8
a4
98.8
5 116
100.0
85
100.0
8 3
Amni
oniti
s/ch
orio
amni
oniti
s v
2.22
(0.10, 51.25)
2 5.
Tot
al
116
100.0
85
100.0
a
-
-
Yes
1
0.9
No
115
99.1
85
100.0
a
Ana
lysi
s res
tric
ted
to c
ases
and
con
trol
s fo
r w
hom
med
ical
reco
rds w
ere
obta
ined
. A
cor
rect
ion
fact
or o
f (0.5) w
as a
dded
to e
ach
cell
whe
n a
zero
cel
l was
enc
ount
ered
in c
alcu
latin
g od
ds r
atio
. *P<0.05.
a
Tab
le 4
. R
isk o
f SI
DS
rela
ted
to a
nten
atal
var
iabl
es, u
sing
live
con
trol
s fo
r co
mpa
riso
n
Ant
enat
al
vari
able
SIDS
Liv
e co
ntro
ls
Odd
s 95
% c
onfid
ence
n
%
n ‘30
ra
tio
inte
rval
Mul
tiple
birt
h Y
es
No
Tot
al
1I2-
mon
th
preg
nanc
y in
terv
al
Yes
N
o T
otal
Pre-
ecla
mps
ia
Yes
N
o T
otal
Ecla
mps
ia
Yes
N
o T
otal
Ute
rine
ble
edin
g Y
es
No
Tot
al
11
119
130 12
93
105 1
129
130 3
127
130 13
11
7 13
0
8.5
91.5
10
0.0
11.4
88
.6
100.
0
0.8
99.2
10
0.0
2.3
97.7
10
0.0
10.0
90
.0
100.
0
1
0.3
29.5
8 32
0 99
.7
321
100.
0
8 3.
3 3.
77
234
96.7
24
2 10
0.0
3 0.
9 0.
82
318
99.1
32
1 10
0.0
- -
17.6
5 32
1 10
0.0
321
100.
0
23
7.2
1.44
29
8 92
.8
321
100.
0
(7.5
6, 1
15.7
5)**
(1.5
7,
9.05
)**
(0.0
8,
7.96
)
(2.0
2, 1
54.6
0).
(0.7
1,
2.93
)
Vagi
nitis
Y
es
No
Tot
al
Sexu
ally
tra
nsm
itted
dis
ease
Y
es
No
Tot
al
Plac
enta
pra
evia
Y
es
No
Tot
al
Abru
ptio
pla
cent
ae
Yes
N
o T
otal
Amni
onit
is/c
hori
oam
nion
itis
Yes
N
o T
otal
6 12
4 13
0.0 5
125
130 4
112
116 6
110
116 1
115
116
4.6
95.4
10
0.0
3.8
96.2
10
0.0
3.4
96.6
10
0.0
5.2
94.8
10
0.0
0.9
99.1
10
0.0
9 2.
8 1.
68
312
97.2
32
1 10
0.0
2 0.
6 6.
38
319
99.3
32
1 10
0.0
21.8
0 -
-
272
100.
0 27
2 10
0.0
4 1.
5 3.
65
268
98.8
27
2 10
0.0
1 0.
4 2.
36
271
99.6
27
2 10
0.0
(0.5
9,
4.77
)
(1.5
0,
27.1
9)*
(2.8
0, 1
69.7
8)**
(1.0
9,
12.2
2)*
(0.1
6,
35.1
3)
23 3 a
2
Ana
lysi
s res
tric
ted
to c
ases
and
con
trol
s fo
r who
m m
edic
al r
ecor
ds w
ere
obta
ined
. A
cor
rect
ion
fact
or o
f (0
.5) w
as a
dded
to e
ach
cell
whe
n a
zero
cel
l was
enc
ount
ered
in c
alcu
latin
g od
ds r
atio
. *P
< 0.
05.
**P<
0.01
. s!
168 G.M. Buck et al. Table 5. Risk of SIDS related to antenatal variables controlling for confounders, using dead controls for comparison
Best model for antenatal variable
Adjusted odds 95 % confidence
Risk factor +[confounders] ratio interval
Multiple birth + [gestation] 1.9 (0.6, 6.2)
Pre-eclampsia 0.8 (0.05, 12.5) Uterine bleeding+ [pregnancy
512-month pregnancy interval 2.4 (0.8, 7.7)
interval +gestation + birthweight] 9.0 (1.2, 71.1). Vaginitis 0.6 (0.2, z.oj Sexually transmitted disease +
[gestation +month prenatal care initiated] ,- 2.6 (0.3, 24.2)
Abruptio placentae +[pregnancy interval +gestation] 2.1 (0.4, 10.6)
P(0.05
Table 6. Risk of SIDS related to antenatal variables controlling for confounders, using live controls for comparison
Best model for Adjusted antenatal variable odds 95% confidence Risk factor +[confounders] ratio interval
Multiple birth+ [gestation] 5 12-month pregnancy interval Pre-eclampsia Uterine bleeding Vaginitis Sexually transmitted disease Abruptio placentae Amnionitis/chorioamnionitis +
[gestation]
19.5 3.8 0.7 1.4 1.6 4.7 1.9
0.8
(2.4, 157.0)* (1.5, 9.6). (0.7, 7.1) (0.7, 2.9) (0.6, 4.7) (0.8, 26.1) (0.7, 5.2)
(0.02, 34.1)
* Px0.05
Uterine bleeding remained a statistically significant risk factor for SIDS using logis tic regression techniques when dead controls represented the comparison group (OR= 9.0). Although the results were not statistically significant, the data suggested that sexually transmitted disease (OR=2.6), 5 1 2 month pregnancy interval (OR=2.4), abruptio placentae (OR=2.1), and multiple birth (OR= 1.9), may be risk factors for SIDS in need of further study.
In utero hypoxia and SZDS 169
Using logistic regression techniques and live controls, multiple birth (OR-19.5) and 512 month pregnancy interval (OR=3.8) were found to be statistically significant risk factors for SIDS. Elevated but not statistically significant odds ratios were observed for sexually transmitted disease (OR- 4.7), abntptio placentae (OR= 1.9), vaginitis (OR= lh), and uterine bleeding (OR--1.4). No statistically significant excess in risk was observed for pre- eclampsia or amnionitis/chorioamnionitis (OR =O. 7 and 0.8, respectively).
Overall, three antenatal risk factors were found to be associated with approximately a twofold or greater increase in SIDS risk using logistic regression: multiple birth (OR=19.5), uterine bleeding (OR=9.0), and 1 1 2 month preg- nancy interval (OR=3.8). The magnitude of risk, however, did vary somewhat for each risk factor depending upon the control group used for comparison.
Discussion
This study was based on 148 autopsy-confirmed SIDS cases, 114 unmatched dead controls and 355 frequency-matched live controls. The findings reported here are unique in that this is a population-based study, the results of which may be generalisable to other Upstate New York live birth cohorts, assuming there is no cohort effect. The reliability of the data is enhanced by high response rates for the collection of matched vital records (97%), hospital delivery records (89%), and a select sample of autopsy reports (100%). As in any study, the data are valid to the extent that data regarding the antenatal variables under study were reliably recorded in medical records. Although the quality of medical records may vary with respect to type of information, there is no reason to assume that the quality of recorded data vaned by infant status which would jeopardise the validity of the study's findings.
The findings reported herein are valid to the extent that misclassification bias with respect to disease status is minimal. One potential source of misclassification bias may be attributed to the incomplete ascertainment of all autopsy reports for SIDS cases and dead controls. The problem is further complicated by the lack of any possible means for reviewing the histology for cases and controls. Our preliminary attempt to estimate the magnitude of misclassifications bias suggested that approximately 5% of the sample was originally misclassified (n=three SIDS cases and one dead control). Unfortunately, the retrospective approach of the study did not lend itself well to histologic assessment of cases and controls.
The data presented here suggest that antenatal factors which are capable of adversely affecting the fetal environment may be associated with an increased risk of SIDS and certainly warrant further study. Univariate analysis identified eight antenatal factors that were apparent risk factors for SIDS, although the results were not always statistically significant across control groups, The eight suggestive antenatal risk factors were: multiple birth, placenta praevia, abruptio
170 G.M. Buck et al. placentae, eclampsia, sexually transmitted disease(s), 5 12 month pregnancy interval, amnionitis/chorioamnionitis, and uterine bleeding. When logistic regres- sion techniques were used, uterine bleeding was the only statistically significant risk factor observed using dead controls for comparison. However, when live controls were used, multiple birth and 112 month pregnancy interval were statistically significant risk factors even after controlling for a variety of potential confounders.
The magnitude of SIDS risk did vary according to control group used for comparison. One possible explanation for the difference in degree of risk by control group may be attributed to some of the methodological limitations of the data, especially the relatively small and fixed number of cases and controls available for analysis. In addition, many of the antenatal factors under study are relatively rare pregnancy-related events. Such small numbers of observed pregnancy events may lead to unstable point estimates and relatively wide confidence intervals, One still cannot rule out the possibility that different results may be related to the unique nature of each control group. In particular, the lack of significant risk factors when using dead controls may be due to the heterogeneous dead control sample or the inclusion of some SIDS cases in the dead control group. Assuming this, it appears possible that the results obtained using the dead controls are underestimated.
The findings presented here are consistent with previous epidemiological research which reported an increased frequency of SIDS associated with a short pregnancy interval,30v3' multiple birth,1a.32*33 history of amnioniti~,)~ uterine bleeding,18*32 illness during p regnan~y ,~~ vaginitis,35 and history of pregnancy cornpli~ations.~~ However, not all earlier studies have reported positive associa- tions between these factors and SIDS. At least two studies failed to observe any association between a positive history of pregnancy complications and SIDS.37J8 Possible reasons for these equivocal findings may be attributed to the methods used for defining and ascertaining cases and controls and the specific types of pregnancy complications under study.
One major shortcoming of this study was the lack of reliable data on maternal cigarette smoking and drug usage (including alcohol) during pregnancy, especially since past research has identified high risk SIDS infants as those born to mothers who smoke cigarettes during pregnancy or narcotic-dependent mother^.^^^ Unfortunately, this study could not assess the effects of smoking or drug usage on risk of SIDS due to a lack of reliable data. Some of the other antenatal variables under study, namely placenta praevia and abruptio placentae, have also been associated with maternal smoking during pregnancy.39 Thus, future studies need to adjust for the possible confounding effects of cigarette smoking and possibly other drug exposures in order to identify which antenatal factors are risk factors for SIDS and which are confounders.
A question of utmost concern to this study is how do these antenatal events
In utero hypoxia and SlDS 171
lead to an increased risk of SIDS? All antenatal variables under study have been associated with a compromised fetal environment, possibly through a diminished or impaired fetal blood flow or oxygen delivery. It remains possible that these adverse antenatal events may act as the initial hypoxic insult to the brain stem. Post-mortem studies have reported that approximateIy 60% of SIDS victims have evidence of chronic hypoxia as indicated by abnormalities in the anatomic sites of ventilatory control and changes in the organs most affected by hypoxia, i.e. brain, pulmonary vasculature, liver, adrenal glands and brown fat.40
A very recent study observed that SIDS victims had prolonged, elevated levels of fetal haemoglobin (F) which was not replaced readily with adult haemoglobin (A).*l The authors suggested that such prolonged, elevated haemoglobin (F) levels may denote a compromised delivery of oxygen to sensitive tissue sites.
In summary, evidence for an increased risk of SIDS associated with in utero hypoxia has been suggested by findings from post-mortem, physiologic and epidemiologic studies. These findings taken together suggest the need to study in utero events in relation to both death and post-mortem findings in order to delineate the relationship, if any, between these factors. Epidemiologic studies have identified possible in utero or perinatal events that may act as the initial hypoxic insult to the developing brain stern structures. Future studies should attempt to determine if these antenatal risk factors are associated with indicators of hypoxia identified at post-mortem analysis. To this end, formal investigation of the hypoxia theory will be possible and, hopefully, offer insight into the natural history of this untoward reproductive outcome.
Acknowledgement
The authors wish to acknowledge the Scholl and National SIDS Foundations for partial support of this study, and the New York State Health Department for their technical assistance.
References
1
2 Gynecology Annuals 1975; 4213-236. 3 Diseases of Children 1982; 136:1012-1023. 4 introduction. Pediatric Annals 1984; 13188-190. 5 of Political Medicine. New York: Praeger, 1986.
Memtt, T.A., Valdes-Dapeila, M. SIDS research update. Pediatric Annals 1984;
Hasselmeyer, E., Hunter, J.C. The sudden infant death syndrome. Obstetrics and
Brooks, J.G. Apnea of infancy and sudden infant death syndrome. American Journal of
Zebal, B.H., Friedman, S.B. Sudden infant death syndrome and infantile. apnea:
Bergman, A.B. The 'Discovery' of Sudden Infant Death Syndrome: Lessons in the Practice
13:193-207.
172 G.M. Buck et al.
6 Shannon, D.C., Kelly, D.H. SIDS and near SIDS. New England Journal of Medicine
7 Peterson, D.R. The sudden infant death syndrome-reassessment of growth retardation in relation to maternal smoking and hypoxia hypothesis. American Journal of Epidemiology 1981; 113:583-589. 8 Chavez, C.J., Ostrea, E.M., Stryker, J.C. et al. Sudden infant death syndrome among infants of drug dependent mothers. Journal of Pediafrics 1979; 95407-409. 9 Rajegowda, B.K., Kandall, S.R., Falciglia, H. Sudden unexpected death in infants of narco tic-dependent mothers. Early Human Development 19 78; 2:2 19-225. 10 Buck, G.M., Cockfair, D.L., Michalek, A.M. et al. Intrauterine growth retardation and risk of sudden infant death syndrome. American Journal of Epidemiology, in press. 11 van Belle, G., Hoffman, H., Peterson, D. Intrauterine growth retardation and the sudden infant death syndrome. In: Sudden Infant Death Syndrome: Risk Factors and Basic Mechanisms, Editors: R.M. Harper L H.J. Hoffman. New York: PMA Publishing Corp., 1988; pp. 203-219. 12 Arsenault, P.S. Matemal and antenatal factors in the risk of sudden infant death syndrome. American Journal of Epidemiology 1980; 111:278-284. 13 Naeye, R.L. Sudden infant death. Scientific American 1980; 24256-62. 14 Valdes-Dapeiia, M., Gillane, M.M., Catherman, R. Brown fat retention in sudden infant death syndrome. Archives of Pathology and Laboratory Medicine 1976; 100547-549. 15 Valdes-Dapeiia, M., Gillane, M.M., Ross, D. et al. Extramedullary hematopoiesis in the liver in sudden infant death syndrome. Archives of Pathology and Laboratory Medicine 1979;
16 Takeshima, S., Armstrong, D., Becker, L.E. et al. Cerebral hypoperfusion in the sudden infant death syndrome? Brain stem gliosis and vasculature. Annals of Neurology 1978;
17 Standfast, S.J., Jereb, S., Janerich, D.T. The epidemiology of sudden infant death syndrome in Upstate New York. Journal of American Medical Association 1979;
18 Standfast, S.J., Jereb, S., Janerich, D.T. The epidemiology of sudden infant death syndrome in Upstate New York, 11. Birth characteristics. American Journal of Public Health
19 Bergman, A.B., Beckwith, J.B., Ray, C.G. Sudden lnfant Death Syndrome: Proceedings of the Second International Conference on the Causes of Sudden Deaths in Infancy. Seattle: University of Washington Press, 1970; pp. 17-18. 20 Villar, J., Belizan, J.M. The timing factor in the pathophysiology of the intrauterine growth retardation syndrome. Obstetrical and Gynecological Survey 1982; 37:499-506. 21 Varma, T.R. Low birth weight babies. The small for gestational age. A review of current management. Obstetrical and Gynecological Survey 1984; 39:616-624. 22 Batzofin, J.H., Fielding, W.L., Friedman, E.A. Effect of vaginal bleeding in early pregnancy on outcome, Obstetrics and Gynecology 1984; 63515-518. 23 Moghissi, K.S. Maternal nutition in pregnancy. Clinical Obstetrics and Gynecology
24 Fedrick J., Adelstein, P. Interpregnancy interval and outcome of pregnancy. British Medical Journal 1973; 4753-756. 25 Ledger, W. Premature rupture of membranes and in utero fetal infection. Clinical Obstetrics and Gynecology 1979; 22:329-337. 26 Chamberlain, G. Epidemiology and aetiology of the preterm baby. Clinics in Obstetrics and Gynecology 1984; 11:297-315.
1982; 306:959-965.
103~5 13-5 15.
4:257-262.
214:1121-1124.
1980; 701061-1067.
1978; 21:297-3 10.
In utero hypaxia and SlDS 173
27 Mantel, N., Haenszel, W. Statistical aspects of the analysis of data from retrospective studies of disease. Journal of National Cancer Institute 1959; 22:719-748. 28 Breslow, N.E., Day, N.E. Statistical methods in cancer research: the analysis of case- control studies. Lyon: International Agency for Research and Cancer (IARC Scientific Publications No. 32), 1980; pp. 192-246. 29 Greenberg, R.S., Kleinbaum, D.G. Mathematical modeling strategies for the analysis of epidemiologic research. Annual Reviews of Public Health 1985; 6:223-245. 30 Lewak, N., van den Berg, B.J., Beckwith, J.B. Sudden infant death syndrome risk factors. Clinical Pediatrics 1979; 18404-411. 31 Standfast, S.J., Jereb, S.K., Aliferis, D.A. et al. The epidemiology of SIDS in Upstate New York. In: Sudden Infant Death Syndrome, Editors: J.T. Tildon, L.M. Roeder & A. Steinschneider. New York Academic Press, 1983, pp. 59-75. 32 Fedrick, J. Sudden unexpected death in infancy in the Oxford Record Linkage Area: details of pregnancy, delivery and abnormalities in the infant. British Journal of Preventive and Social Medicine 1974; 28164-171. 33 Kraus, J.F., Borhani, N.O. Post-natal sudden unexplained death in California- a cohort study. American Journal of Epidemiology 1972; 95492-510. 34 Naeye, R.L. Intrauterine septicemia: recognized cause of hypoxia and ischemia responsible for altered brain stem structure and function. Biology of the Neonate 1977;
35 Naeye, R.L., Ladis, B., Drage, J.S. Sudden infant death syndrome. American Journal of Diseases of Children 1976; 1301207-1210. 36 Murphy, J.F., Newcombe, R.G., Sibert, J.R. The epidemiology of sudden infant death syndrome. Journal of Epidemiology and Community Health 1982; 36:17-21. 37 Froggatt, P., Lynas, M.A., Marshall, T.K. Sudden death in babies: epidemiology. The American Journal of Cardiology 1968; 22457-468. 38 Protestos, C.D., Carpenter, R.G., McWeeny, P.M. et al. Obstetrics and perinatal histories of children who died unexpectedly ('Cot Death'). Archives of Disease in Childhood
39 Naeye, R.L. Placenta previa: predisposing factors and effects on the fetus and surviving infants. Obstetrics and Gynecology 1978; 52521-525. 40 Kelly, D.H., Shannon, D.C. Epidemiology clues to the etiology of SIDS (editorial). American Journal of Public Health 1980; 701047-1048. 41 Giulian, G.G., Gilbert, E.F., Moss, R.L. Elevated fetal hemoglobin levels in sudden infant death syndrome. New England Journal of Medicine 1987; 3161122-1 126.
32189-192.
1973; 48:835-841.