An Epidemiologic Study of Contraception and Preeclampsia Hillary S Klonoff Cohen, PhD; David A Savitz, PhD; Robert C Cefalo, MD, PhD; Margaret F McCann, PhD The primary hypothesis of this study was th.
Trang 1An Epidemiologic Study of
HillaryS.Klonoff-Cohen, PhD;DavidA.Savitz, PhD;Robert C.Cefalo, MD, PhD; MargaretF.McCann,PhD
The primary hypothesis of this study was that contraceptive methods that
prevent exposure to sperm and seminalfluid (condoms, diaphragms,
spermi-cides, withdrawal) are associated with an increased risk of developing
pre-eclampsia duringthesubsequentpregnancy.A case-controlstudywas
conduct-ed comparingthe contraceptive and reproductive histories of110 primiparous
women with preeclampsia with 115 pregnant women without preeclampsia,
aged15to35years, who gave birthatNorth CarolinaMemorialHospital, Chapel
Hill,between1984and1987.Controlswerefrequencymatchedtocasesbyage,
race, anddistance from thehospital. Unconditional logistic regression analysis
indicated a2.37-fold (95%confidence interval, 1.01 to 5.58) increased riskof
preeclampsia for users of contraceptives that prevent exposure to sperm A
dose-response gradientwasobserved,with increasingrisk ofpreeclampsiafor
thosewithfewerepisodesof sperm exposure These resultsweresupportiveof
thehypothesisthat birth control methods thatpreventsperm exposure mayplay
arole intheetiologyofpreeclampsia.
(JAMA.1989;262:3143-3147)
themostimportant unsolvedproblems
in obstetrics.1 It is the third leading
causeof maternalmorbidityandamajor
cause of intrauterine growth retarda¬
tion and perinatal morbidityandmor¬
tality.2
For editorialcomment seep3184.
Theetiologyofpreeclampsiaislarge¬
ly unknown Genetic predisposition,35
prostacyclins,6 environmental factors
such as a virus7 or solvents,8 and the
immunesystem9"" have all been
impli-cated, although none completely ex¬
plainsthe etiologyof the disease Any plausible hypothesishastoexplain why (1) preeclampsiais farmore commonin primigrávidas12; (2)withafirstpregnan¬
cyofatleast 37 weeks'duration,the risk
of preeclampsia is reduced in subse¬
quent pregnancieswith thesamepart¬
ner13; and (3) there appears to be an
increased incidence of preeclampsia
with paternal change in multiparous women,14"16 although this has recently
beenchallenged bya casestudy.7 Several studies have considered a
link betweenpaternalfactors andpre¬
eclampsia. An elevated incidence of
preeclampsiawasfound among artificial donor inseminationpregnanciesforpri¬
migrávidas (10%, expected value 5%)
andmultigravidas (7.8%, expectedval¬
ue 0.9%), reflecting the effect of a
changed paternity.17Inanoocytedona¬
tion study,18 5 of the first 10 patients
treated forinfertilitywithoocyte
dona-tion developed preeclampsia. In addi¬
tion, Marti and Harrmann19 investi¬
gated exposure to paternal spermatic antigenstodetermine if continuous and
regular exposure ofthe female genital
tract to spermreduced theincidenceof
preeclampsia by examining the use of oral contraceptives among 28
pre-eclamptic women. Theyfound that the controlgrouphad three timesmoreepi¬
sodes ofunprotectedsexual intercourse than thecase group, supportingapro¬ tective effect of sperm contacts. The
studywaslimitedby overly restrictive exclusion criteria and afailure to note theuseof othertypesof birth controlby
85%(24/28)of thecases.Inaddition,the time period overwhich exposure was
considered was not stated Nonethe¬
less, this was the first study, to our
knowledge, toquantify the amount of sperm exposureandtoconsiderapossi¬
ble connection between spermatozoal histocompatibility and preeclampsia.
No other studiestodate have focusedon
the relationship between all types of birth control andpreeclampsia.
Toextend thesefindings, a case-con¬
trol study was designed to determine whether therewas anincreased risk of
preeclampsia among women who used methods of birth control that prevent
exposure ofthe endometriumtoejacu¬
late(eg,barriermethods, spermicides) comparedwithwomenwhodidnot use
thosemethods(eg,usersof intrauterine
devices, oral contraceptives, rhythm).
With the nonbarrier methods, sperm
comesincontactwith theendometrium,
where it is absorbed With the barrier
methods,nosperm(or onlydeadsperm)
comeincontactwith theendometrium,
and the maternal immunesystemispre¬
sumably notexposed topaternal anti¬ gens.The total number ofreported
con-From the Department ofEpidemiology, School of
Public Health (Drs Klonoff-Cohen, Savitz, and
McCann),and the Department of Obstetrics and
Gyne-cology,Division of Maternal and Fetal Medicine, School
of Medicine (DrCefalo),Universityof North Carolina at
ChapelHill Dr Klonoff-Cohen is now with the
Depart-ment ofCommunityand FamilyMedicine, School of
Medicine,Universityof California\p=m-\SanDiego,La Jolla.
Reprint requests to 79680 Bermuda Dunes Dr,
Ber-muda Dunes, CA 92201 (Dr Klonoff-Cohen).
Trang 2tactswith spermwasalso evaluatedto
determine iffewer episodes of unpro¬
tected intercourse (resulting in fewer
sperm contacts) increased the risk of
preeclampsia.
METHODS
SubjectSelection
Cases and controls were selected
from female residents of North Carolina
aged 15to35yearswhowerehospital¬
izedattheUniversityof North Carolina
Memorial Hospital, Chapel Hill, fora
live birth of 26 to 42 weeks' gestation
betweenJanuary 1, 1984, andJanuary
30,1987.This timeperiodwaschosento
yield approximately 175cases and 175
controls,which wouldprovidesufficient
statisticalpower(90%) todetectarisk
ratio of 2.0orgreaterfor barriercontra¬
ceptionuse.17Alleligible subjectswere
primiparous, includingasmall group of
multigravidwomenwhohadaprevious
first-trimesterspontaneousorinduced
abortion Multiparous women with
preeclampsia were omitted since they
are morelikelytohavesomeunderlying
disorder suchasocculthypertensionor
renal disease that isresponsiblefor the
increased bloodpressure.
Cases ofpreeclampsia were initially
identified through review of the com¬
puterized discharge diagnoses,withre¬
view of medical records to determine
eligibility. Over a 3x/2-year ascertain¬
ment period, 420 potential cases were
identified, amongwhom 64% were ex¬
cluded Thereasonsfor exclusionwere
age of less than15or morethan 35years
(20%); multiparity (20%); the presence
ofpreexisting diseases,suchasdiabetes
or hypertension (20%); and incorrect
data in the chart(4%).This left 150eligi¬
blecases.
Eligible controls were women who
had a live birth during the same
3V2-year period. There were 2100 eligible
controls, of whom 420 (20%) were se¬
lected Among potential controls, the
exclusion rate was almost identical to
thatamong the cases: 30% because of
thepresenceof otherdiseases,20% be¬
causeofmultiparity,and10%because of
age, resultingin 150 eligible subjects.
Controls were frequency matched to
cases by age (in 5-year strata), race
(white or black), and geographic loca¬
tion(distancefromthehospital).
Information on demographic vari¬
ables, medical history, antepartum
problems, and labor and delivery was
obtained through medical record re¬
view Written consent was obtained
from the physician and then from the
patient to participate in a 15-minute
telephone interview The interview
elicited information about sexual
con-Table1.—DiagnosticCriteria*
MildPreeclampsia
All of thefollowing signsafter 20 weeks ofgestation:
1 An increase insystolicpressure to 140 mmHgor an increase in systolic pressure of >30 mmHgabove usual,
taken 2 times 6 hours apart at bed rest
2 An increase in diastolic pressure to 90 mm Hg or an Increase in diastolic pressure of 15 mm Hg more than
usual, taken 2 times 6 hours apart at bed rest
3 Presence of 2300 mg of protein In clean-catch urine in at least 2 random urine specimens collected 6 hours
apart
4 Edema of the face or hands of >1 + or a gain of >5 lb In 1 week
SeverePreeclampsia
One or more of thefollowingsigns:
1. Systolicpressure of 160 mm Hg or diastolic pressure of 110 mm Hg recorded 2 times 6 hours apart with
patient at bed rest
2 Proteinuria, 5 g in 24 hours, or a 3 to 4+ protein on dipstick
3. Oligurla,urinary output of <400 ml_ In 24 hours
4 Cerebral or visual disturbances, including eye changes
5 Pulmonary edema or cyanosis
6. Epigastricpain
7 Evidence of hemolysls, abnormal result from liver function test, and falling platelet count (HELLP syndrome)
Eclampsia
Generalized convulsions and/or coma
From The AmericanCollegeof Obstetrics andGynecology20as modified for this study.'
tactwith the father of the child of the indexpregnancy(includingall forms of
contraception,the timeperiodof sexual
relations, frequency of sexual inter¬
course, and thelengthof time to preg¬
nancy without contraceptive use) as
well aswith all other sexualpartners.
Thequestionnairewasdesignedina se¬
quentiallifehistory approachtorecon¬
structthereproductiveandcontracep¬
tive practices usinglife events Single
womenwhoweresporadicusersof birth controlorusednobirth controlrequired
aspecial sequence ofquestions Ques¬
tions were also asked about personal habits, including the use of alcohol, medications, andcigarettesbefore and
duringpregnancy
A 10% sample was reinterviewed 1 month after the initial interviewtoas¬
sess validity Also, the questionnaire
datawere comparedwith the informa¬
tion obtained in the sexual/contracep¬
tivehistory takenby a nurse orsocial worker and recorded in the medicalrec¬
ordsatthe time of thepregnancy.This
historyincluded the age at menarche,
age atfirstintercourse,number ofsexu¬
alpartners, andtypes and duration of birth control This studywasapproved
bythe Human Subjects' Committeeat theUniversityof North Carolina School
of Medicine
DiagnosticCriteria forPreeclampsia
Other studies of preeclampsia are
plaguedwithmisdiagnoses.The criteria usedtodiagnose patientswereadopted
from the AmericanCollegeof Obstetri¬
cians andGynecologists20 and modified for thisstudy (Table l).121The fullspec¬
trum of severity of preeclampsia was
included, with the adherence to these
specific criteriaensuringthateventhe mild cases were indeed preeclampsia.
Thediagnosticcriteriawere asfollows:
(1) the triad ofchanges in blood
près-sure, proteinuria, and edema must all have beenpresentat20weeksorlater
ingestation; (2)thesubjectsmusthave hadacomparativeprepregnancy base¬ line blood pressure measurement re¬
corded in thechart;and(3)aminimum
offourreadingsofbloodpressure,urine
protein excretion,andweight gaindur¬
ingpregnancymust have been record¬
ed Toensurecomparabilityinprenatal
care, controls were also required to have had these data available
Cases and controlswereexcluded for the following reasons: antecedent or
concurrentdiseases such asrenal, car¬
diac,orvascular disorders(preexisting hypertension) and endocrine diseases
(diabetes mellitus); multiple pregnan¬ cies and hydatidiform moles; and age less than 15 orgreaterthan 35 years The phenomenon of intrapartum and
postpartum (transient) hypertensionof laborwasalsoexcluded inthisstudy.
ClassificationofExposure
Barriermethodswerethose that lim¬
it the travelofspermatozoaatthe cer¬
vix (cervical caps, sponges, and dia¬
phragms) and in the vagina (spermi-cides, which denature the sperm; and
condoms), thus preventing contact of
living sperm antigens with the endo-metrial lining. Withdrawalwas classi¬ fied as abarrier method, with the as¬
sumption that no sperm entered the
vaginal vault With the nonbarrier methods (oral contraceptives, intra-uterinedevices, rhythm, andnonusers
of any method), the spermatozoa are
freeto travel from thevagina through
the cervical canalto the uterinecavity
and fallopian tubes Only sexual rela¬ tions with the baby'sfather were con¬
sideredintheclassification ofexposure. Fora womantobe classifiedas abarrier methoduser, shemusthave usedabar¬ rier method exclusively for the entire
Trang 3time of her sexual relations with the
baby'sfather
Ifa woman used both a nonbarrier
and barrier method consecutively, she
was classified as a non-barrier user,
since the woman's endometrium would
have beenexposedto the spermof the
baby'sfatherpriortoconception.
The number ofcontactsbyspermwas
quantified by multiplyingthefrequency
of intercourse without barrier methods
(per month) by the duration ofuse (in
months).19 Exclusive barrier users
would contribute zero contacts and
were therefore included in the lowest
dose category. Barrier method users
who had months duringwhichtheydid
not use barrier methods accumulated
exposuresduringthose months(which
wouldincludethe interval betweences¬
sation of contraception and time of
conception).
These computations included
primi-gravidandmultigravidwomen(n=47).
For primigravid patients, the number
ofcontactswasbasedsolelyonthesexu¬
alhistorywith the father of the child If
the multigravid patientshad thesame
partnerfor bothpregnancies, then the
number ofcontactswascombined If the
partner was different for the abortion
andsubsequent delivery,thenonly the
number ofcontacts with the father of
the child from the second pregnancy
wasincluded
StatisticalAnalysis
Thecrudeodds ratiorelatingbarrier
contraceptiveuse topreeclampsiawas
calculated, followedbystratifiedanaly¬
sistoderive Mantel-Haenszeladjusted
oddsratios.22Unconditionallogisticre¬
gressionwas used to assess the expo¬
sure-disease relationship while simul¬
taneously considering interaction and
confounding.22Datawereanalyzedwith
the Statistical Analysis System soft¬
ware(SASInstituteInc, Cary, NC).
RESULTS
Characteristics ofSubjectsand
ResponseRate
Among 300 eligible subjects, 75%
wereinterviewed(73%ofcasesand77%
ofcontrols). Losswasprimarilycaused
by the inability to locate respondents
(26%ofcasesand 22% ofcontrols),with
a small number ofrespondentrefusals
(0.7% of cases and 1.3% of controls).
There were tremendous difficulties in
locatingandinterviewing patientsfrom
thispredominantlyruralpopulationbe¬
cause of their transient life-style and
lower socioeconomicstatus.This result¬
ed in the lack ofregular employment,an
absence oftelephones,noprivate physi¬
cians, and the hesitationbyfriends and
relativestodivulge telephonenumbers
Table 2.—Characteristics of the InterviewedPopulation
Dichotomous Variable
Case Persons PersonsControl
No % No %
Odds Ratio
Total 95% Confidence Interval
1 Pregnancy
2Pregnancies
Biologic/lmmunologicVariables
89 83 80 71 ]
18 17 32 29 1.04-3.78
Familyhistory of preeclampsia
Nohistoryof preeclampsia
29
78 27 73
5
107 96 7.96 3.30-19.23
Age at menarche, y
£13
>13
Ageat first sex, y
==17
>17
Reproductive Variables 72
35 67
33
87 24 78
22 0.57 0,31-1.04
No of partners
£2
>2
63 59 41
67 45 60
40 0.96 0.56-1.65
Age at pregnancy, y*
£21
>21
61 44
58 42 65 45
59
47
60 44 56
55
57
49
51 0.81 0.48-1.38
Education, y
£12
>12
41
21
Demographics
66 34
46 35
57
43 1.49 0.75-2.95
Working duringpregnancy
Not workingduringpregnancy
70
37 65 35
61
51 54
46 1.58 0.92-2.73 White
Black
64
40
63 49 56
44 1.16 0.68-1.98 Married at time of pregnancy
Singleat time of pregnancy
51
56
48 52
44
68
39
61 0.82-2.41 Prenatal visit 1
£13 wk
>13 wk
Other Variables
56
37 60 40
56 52
52
48 Total No of prenatal visits
£10
>10
50
42 54 46
52 54 49
51 1.24 0.71-2.17
Hypertension in subject's mother
No history of maternal hypertension
46
61 43 57
36 76 32
68 1.59 0.92-2.76
Hypertension In subject's father
No history of paternalhypertension
34 73 32 68
21
91 19
Smokingduring pregnancy
Nosmoking duringpregnancy
33
74 31 69
38 74
34
66 0.87 0.49-1.53 Alcoholduringpregnancy
No alcoholduringpregnancy
32
75 30 70 37
75
33
Infant weight, g
£3200
>3200
PregnancyOutcomes 64
39
42
70
37
63 2.60 1.51-4.48
Placentalweight,g
£590
>590
60
41 47
58 45
55 1.76 1.02-3.06
Gestational age, wk
£39
>39
69
36
44
*Mean age at pregnancy ( ±SD)was 21.69 ( :
persons 4.83) years for case persons and 22.13 ( ± 5.15) years for control
becauseofsomeof thesubjects' hospital
debts Manyofthesewomen were sin¬
gleatthe time of pregnancyordelivery
butsubsequentlydidmarryand there¬
fore had differentsurnamesatthe time
ofthe interview
Nonrespondentswerecomparedwith
respondents with respect to the only
data available fornonrespondents:race, age, and prenatal care. There was a
highernonresponse rate among black
cases(34%) comparedwith whitecases
(23%), although white(27%) and black
(30%)controls hadaverysimilar
nonre-sponse rate.Youngercases(=s21 years)
had a higher nonresponse (39%) than older cases (24%); however, nonre¬
sponse among controls was similar
across age groups The nonresponse rate by number ofprenatal care visits
wassimilar for bothcasesandcontrols Even though there were some differ¬
encesinresponse, adjustmentforrace
and age should eliminateanybias. The characteristics of the case and controlrespondentsaresummarized in Table 2 Cases and controlswerevery similar withrespectto race,age, alcohol
Trang 4consumption, smoking history, educa¬
tion,ageatmenarche,age atfirst inter¬
course,and number of sexualpartners.
Caseswere morelikelythan controlsto
reportafamily historyofpreeclampsia,
lower infant weight, lower placental
weight,shortergestationalage,having
worked during pregnancy, and fewer
prenatalvisits
Sixty percentofcases werewhite and
40%wereblack(Table 2), derived from
an obstetric population of30% whites
and70%blacks.Accordingtothe litera¬
ture, preeclampsiaistypicallyadisease
withahigherincidence inblacks,23but
thiswasnotapparentinthisstudy.This
mayhave been the result ofmoreblacks
than whites being excluded from the
study primarily because ofunderlying
hypertension andgreaternonresponse
among black cases. The mean age for
both cases and controls was 22 years
Although subjects between 15 and 35
years of age were included, only 10%
wereless than18 years oldor overthe
ageof29 years.Approximatelyhalf the
cases were not married and half had
fewer than 12yearsof education Over
onefourth of thecasesreportedafamily
historyofpreeclampsia.
ContraceptiveUse and
Preeclampsia
Twenty percent of the 110cases(21
women)andonly9% of the 115 controls
(10 women)used barrier methods exclu¬
sively. This resulted in a crude odds
ratio of 2.48 (95% confidence interval,
1.13 to 5.49), comparing barrier with
nonbarriercontraceptiveusers.
Eight preeclamptic women (but no
controls)had intercourseonlyonce,us¬
ingnoform of birth control, andwere
placedinthe nonbarriercategory.How¬
ever,they technicallyhadnotbeenpre¬
viously exposedto any spermfrom their
partners, so, theoretically, they belong
in the barriergroup.Thischangewould
result in an even larger odds ratio of
3.76foruseof barriercontraceptives.
Amongboth cases andcontrols, the
most frequent birth control practice
usedduringintercourse with the father
of the childpriortotheir firstpregnancy
was no birth control; the second most
common response was oral contracep¬
tives Thegreateruseofbarrier meth¬
odsamongcases waslargelyaccounted
for by increased use of condoms (11
cases,0controls).
Fourcategoriesofpossible
confound-ersof therelationshipbetween the bar¬
rier method of birth control and pre¬
eclampsia were considered: biological,
reproductive, demographic, and other
suspectedvariables(Table 2).
The variableswereconsideredoneat
a time in stratified analysis, and the
adjusted odds ratio for each variable
wascomparedwith the crude odds ratio
of 2.48 There was little indication of
confounding (adjusted odds ratios
ranged from 2.40 to 2.60), except by gravidity (adjusted odds ratio, 2.71),
marital status (adjusted odds ratio, 2.24), and paternity status (adjusted
odds ratio, 2.97) However, the confi¬
dence intervals for these three oddsra¬
tioswereverywide. Furthermore,the calculation ofanadjusted odds ratio is
questionablewhen interaction is pres¬
entand therearesmall numbers insome
cells.22
Stratum-specific odds ratios were
comparedwith the crude odds ratioto
assess effect modification There was
someindicationthat the odds ratio for
use of barrier methods was especially
elevated(oddsratio3=3.4)in the follow¬
ing strata: single mothers, first preg¬
nancy, family history ofpreeclampsia, hypertensioninthesubject'smotheror
father, less than12 years ofeducation,
andnotworking duringpregnancy
Logistic Regression Analysis
In the unconditional logistic reg¬
ressionmodelrelating typeofbirthcon¬
trol to preeclampsia, seven covariates
thoughttobe themostimportant poten¬
tial confounders or effect modifiers
were included: gravidity, marital sta¬
tus, smoking duringpregnancy, alcohol
consumption duringpregnancy, family historyofpreeclampsia, workingstatus
duringpregnancy, and ahistoryofhy¬
pertension in the subject's mother
Race, age, and geographic location
were frequency matched Their inclu¬
sion inaconditionallogistic regression
model did notaffect the odds ratio, so
thatonlyunconditionallogisticregres¬
sion resultsarereported.
Only interpretable two-way interac¬
tions were considered for inclusion in the model Theonly significantinterac¬
tion was contraceptive exposure by
maritalstatus,whichhadnotbeensug¬
gestedinthe literature Therefore, lo¬
gistic analysis wasconsidered both in¬
cluding and omitting the interaction term
Theadjustedodds ratio for the final
model, containingthe abovesevenmain effects and no interaction terms, was
2.37 (95% confidence interval, 1.01 to
5.58). The adjusted odds ratio for the final model containing the interaction termforexposurebymaritalstatusand thesevenmain effectswas0.39(0.26to
0.59)for marriedwomenand 26.08(8.50
to79.84)forsinglewomen. Thoughim¬
precise, there is evidence of an in¬
creased risk ofpreeclampsiaassociated with barriercontraceptives onlyamong
singlewomen.
Amount ofSperm Exposureand
Changed Paternity
Adose-responseeffectwasobserved when the total number of sperm con¬
tactswasanalyzed, withcategorieses¬
tablishedby dividingthestudy popula¬
tion (107 cases and 112 controls) into
quartiles by total number ofcontacts
Compared with women with 480 or
more contacts, intervals of 181 to479, seventy-threeto180, andfewer than 73 contacts produced odds ratios of 1.34, 1.80, and2.41, respectively. Thelogis¬
ticregression equationresultssuggest increasingriskofpreeclampsiawith de¬
creasingamountsofspermand seminal fluid exposure by a factor of1.34 per
quartile.
The effect ofchanged paternitywas
evaluated in the smallsubpopulationof
multigravidas (n=47). Whenapartner
other than the fatherwasinvolvedin a
previousinduced orspontaneousabor¬
tion,the crude odds ratiorelatingbarri¬
ercontraceptiveusetotherisk ofpre¬
eclampsia was 2.34 (95% confidence
interval,0.68 to8.03).Atrendwasalso observed relating amount ofexposure with the father of the childto
preclamp-siarisk,withcategoriesestablishedby dividing this subpopulation into quar¬ tiles This trend wassimilar inmagni¬
tude to thatobserved for all subjects Compared with women with 644 or
morecontacts,intervals of 288to643,
onehundredfifty-oneto287,and fewer than 151 contactsproducedodds ratios
of1.40,1.96,and2.74, respectively.
COMMENT The results of this case-controlstudy
indicate thatwomenwho used barrier
contraceptiveswere overtwiceaslikely
as womenwho used nonbarriercontra¬
ceptivesto develop preeclampsia. Fur¬
thermore,womenwhowereexposedto smaller amounts of sperm were at
greaterrisk ofpreeclampsia.
An interaction of contraception by
maritalstatuswasnoted,withanodds ratio formarriedwomenusingbarrier methods of0.39, comparedwithanodds ratio forsinglewomenof 26.18 As indi¬ catedbythe wide confidenceintervals,
this result couldpossiblyreflect random
error. Alternatively, different factors
might affect preeclampsia in married
women compared with single women.
The irregularity of intercourse among unmarriedwomen,differinginfluences
inchoice of birth controlbymaritalsta¬
tus, varyinglevels ofcareinuseofcon¬
traceptives,or someotherconfounders nottaken intoaccountarepotentialex¬
planatoryfactors that needtobe inves¬
tigatedinfuture studies Apossibleex¬
planationfor the differences in the odds
Trang 5ratio for married andsinglewomenmay
be the"slippage" thatoccurswhenus¬
ing barrier methods Since married
womentypicallyhaveonesteady part¬
nerwith regular intercourse, the slip¬
pagethatoccursismostlikelywith the
father of thechild, thereby exposingthe
endometrium to the same paternal
spermantigens morefrequently. With
unmarriedwomen,however,theremay
be several partners involved, withsex
on anirregularbasis.Thus,therewould
be lessopportunityfor inadvertentex¬
posure tothefather'ssperm
Preeclampsia may be influenced by
factors suchas amaternal inheritance of
a predispositionto preeclampsia,4gra¬
vidity,1012 and a change ofpartner14as
wellasexposure to spermand seminal
fluid through intercourse Several
mechanisms might explainan effect of
the latter: trophoblast-lymphocyte
cross-reactive antigensfrom the semi¬
nalplasma,whendepositedin thevagi¬
na,couldperhapsserve as anantigenic
sourceofallogeneic recognitioninwom¬
en24; an unidentified agent might be
presentinthe spermorseminalfluid;or
certain forms of birth control through
nonimmunologic mechanisms might
preventnormaladaptivemechanismsin
pregnancy
While thepercentageofwomenwho
had used barrier methods exclusively
was higher among cases, there were
many cases who had used nonbarrier
methods This latter observation isnot
surprising giventhe multifactorial etiol¬
ogy of this condition. Furthermore,
amongthecases who were nonbarrier
users,there is alargegroup who have
also used barrier methods forvarying
amountsof time.Perhapstheamountof exposure without barrier method use
was not sufficient to protect from
preeclampsia.
As in any telephone survey, the na¬
ture of the questions, length of recall
required,andthe interviewsettingmay all affect the accuracy ofthe informa¬
tion Emphasis wasplaced onthe reli¬
abilityandvalidity25of thepersonalin¬
terviews,since all thecriticalexposure information was gathered in thisway.
Theagreementratefor the 10%sample
thatwasreinterviewed wasextremely high for contraceptive history (com¬
pleteconcordance for 90% ofcasesand 89% ofcontrols). The agreement rate between the medical records and pa¬
tient'shistoryfortypesof birth control used was also excellent (100%). The number of months ofuseforeachtypeof birthcontrolwasverified in the medical records26,27 and found to be in exact
agreementfor 86% ofcases and84%of controls Anyerrors areunlikelytobe different between cases and controls,
because there is no publicly perceived
association between preeclampsia and birth controlmethods, sotheresulting
biaswould be toward the null
Another potential limitation ad¬
dressedwithin thisstudywasselection bias TheUniversityofNorthCarolina MemorialHospitalisahigh-riskobstet¬
ric center fora large catchment area, which could result in the referral of
casesfromawidergeographicareathan controls Therefore, controlswere fre¬
quencymatchedtocases ongeographic
locationtominimize thisproblem.
Although this study is relatively large compared with previous studies
ontheetiologyofpreeclampsia,theac¬
tual studysize is small, and therefore the study should be replicated with a
muchlarger sample. Thisappliesespe¬
ciallyto the apparentinteraction with marital status Finally, regional pecu¬ liarities in the choice of birth control methods ofcasesand controls because of
education, religion,orculture may limit
generalizabilitytoothersettings.
Thisstudyis consistent with the find¬
ings of Marti and Harrmann19ofa de¬ creased risk ofpreeclampsia with the
use of oralcontraceptives. It mightbe
speculatedthatwomenwithhyperten¬
sion would notuse oral contraceptives
and thereforewould be morelikely to usebarrier methods However,women
with a history of hypertension were
omitted from the study, andtheuse of oral contraceptives was similar for
casesandcontrols
The overall association between bar¬ rier methodsandpreeclampsia (oddsra¬
tio, 2.37) wouldhave important public
health consequences if therelationship
proves tobe causal Thisstudy suggests
that barrier methodsmaycontributeto
asmuchas 60% ofpreeclampticcases.
However, additionalstudiesareneeded before appropriate recommendations
can be made regarding birth control methods based on their risk of pre¬
eclampsia.
We are indebted toRandyCohen, MD, for his
expertadvice,DavidKleinbaum, PhD,and Joanne
Mills, MSPH,for their statisticalassistance,and Deborah Schmidt for her technicalprowess. References
1 PritchardJA, MacDonaldPC, GantNF, eds.
Williams Obstetrics 17th ed EastNorwalk,Conn:
Appleton&Lange;1985.
2. GarreyM Ostehiem Illustrated 2nd ed New
York,NY: ChurchillLivingstone;1982.
3 SutherlandA, CooperD, HowieP, ListonW,
MacGillvrayI The incidence of severe
preeclamp-sia among mothers and mothers-in-law of
pre-eclamptics and controls Br J ObstetGynaecol
1981;88:785-791
4. ChesleyL,CooperD Genetics of hypertension
in pregnancy:possible singlegene control of
pre-eclampsiain the descendants ofeclampticwomen.
Br J ObstetGynaecol.1986;93:898-908
5. CooperD Genetic control ofsusceptibilityto
eclampsiaandmiscarriage.Br J ObstetGynaecol
1988;95:644-653
6 O'BrienW, KnuppelR,SabaH,AngelJ,Benoit
B,Bruce A Serumprostacyclinbindingand half\x=req-\
life in normal andhypertensivepregnantwomen.
ObstetGynecol.1989;73:43-46
7. KilpatrickD,JazwinskaE,ListonW,Smart G.
Severepreeclampsiainmultiparouswomen:
indica-tion of an environmentaltriggering agent?Scott
Med J.1987;32:8-10
8 Eskenazi B, Bracken M, HolfordT, GradyJ.
Exposuretoorganicsolvents andhypertensive
dis-orders of pregnancy Am J Ind Med. 1988;14:177\x=req-\
188.
9 Beer A Possible immunologic bases of
pre-eclampsia/eclampsia.Semin Perinatol. 1978;2:39\x=req-\
56.
10 Beer A, Need J. Immunological aspects of
preeclampsia/eclampsia Birth Defects 1985;
21:131-154.
11 ScottJ,Beer A.Immunologicaspectsof
pre-eclampsia.Am J ObstetGynecol.1976;125:418-425
12. MacGillvrayI Some observations on the inci-dence ofpreeclampsia.J ObstetGynaecolBr
Em-pire.1958;65:536-539
13.CampbellD,MacGillvrayI,Carr-Hill P
Pre-eclampsiain second pregnancy Br J Obstet
Gynae-col.1985;92:131-140
14 Need J.Preeclampsiain pregnancyby
differ-ent fathers Br Med J.1975;1:548-549
15 Ikedife D.Eclampsiainmultipara.Br Med J.
1980;5:985-986
16. FeeneyJ,Scott J. Preeclampsiaandchanged paternity Br J Obstet Gynecol Reprod Biol.
1980;11:35-38
17 NeedJ,BellB,Meffin E.Preeclampsiain
preg-nancies from donor inseminations JReprod Im-munol.1983;5:329-338
18 SerhalP,Craft I Immune basis for
preeclamp-sia: evidence from oocyte recipients Lancet.
1987;1:744
19 MartiJ,Harrmann H.Immunogestosis:a new
etiologicconceptof essential EPHgestosis, with
specialconsideration of theprimigravid patient
Am J ObstetGynecol.1977;128:489-493
20. Management of Preeclampsia Washington,
DC: AmericanCollegeof Obstetricians and Gyne-cologists;1986:19 Technical bulletin.
21. QueenanJ, Hobbins J ProtocolsforHigh-Risk Pregnancies Oradell, NJ: Medical Economics
Books;1983.
22 KleinbaumD,KupperL,MorgensternH. Epi-demiologicResearch:PrinciplesandQuantitative
Methods.Belmont,Calif: LifetimeLearning
Pub-lishers;1982.
23. Chesley L. HypertensiveDisorders in
Preg-nancy EastNorwalk,Conn:Appleton&Lange;
1978.
24. Kajino T, TorreyJ, McIntyre A, Faulk W.
Trophoblast antigens in human seminal plasma
Am JReprodImmunol.1988;17:91-95
25. UdryR.Frequencyof intercourseby daysof the week J Sex Res.1970;6:229-234
26. TilleyB Acomparisonof pregnancyhistory
recall and medical records Am J Epidemiol 1985;121:269-281
27. RosenbergM. Agreementbetween womens'
historyof oral contraceptiveuse andphysicians'
records Am JEpidemiol.1983;12:84-87