A Multivariate Analysis of Risk Factors for Preeclampsia

Brenda Eskenazi, PhD; Laura Fenster, MPH, PhD;

Stephen Sidney, MD, MPH

risk factors for preeclampsia remain un clear, however, and thus preeclampsia has been called a "disease of theories."6 With the exception of nulliparity (ap proximately 75% of women with pre eclampsia are nulliparous)7 and a previ ous history of preeclampsia in multpa ras,8 few other risk factors are univer it has been suggested that the risk fac tors for and causes of preeclampsia in nulliparas and multparas may be differ ent,9"" few have examined these differences. Some researchers have speculated that preeclampsia is a disease of the upper class,12 others more recently be lieve it is a disease of the impover ished,613 and still others think all social classes are at equal risk.14 It has been difficult to determine the exact role of socioeconomic status, since most stud ies have not separated out the contribu tions of socioeconomic status-related variables such as race, employment sta tus, and nutritional status. Nor have these studies controlled for or excluded women with preexisting diseases that may be related to preeclampsia, such as essential hypertension, cardiovascular disease, and diabetes, and that may be more prevalent in groups with lower socioeconomic status. Black women have been observed to be at greater risk for developing pre eclampsia. Chesley et al9 caution that this increased risk of preeclampsia may be due to the higher prevalence of pre existing hypertension in black women. Nevertheless, the Collaborative Peri natal Project15 found a higher incidence of both preeclampsia and eclampsia in black women compared with white women whether or not they had preex isting hypertensive disease. Extremes of age are also postulated as risks for preeclampsia16; however, the effect of age may be confounded by other factors. For example, younger women are more likely to be primigrvi das, and older women are more likely to have essential hypertension.'

Setting.\p=m-\Womenwho gave birth at Northern California Kaiser Permanente Medical Centers in 1984 and 1985. Participants.\p=m-\Preeclampticcases (n =139) were determined from discharge diagnosis of severe preeclampsia and by confirmation of blood pressures and proteinuria from medical records. Controls (n 132) were randomly selected women who had no discharge diagnosis of any hypertensive disorder of pregnancy and who had no evidence of hypertension or proteinuria from medical record review. Main Variables Examined.\p=m-\Medicalrecords were abstracted for information regarding maternal age, race, previous pregnancy history, family medical history, socioeconomic status, employment during pregnancy, body mass, and smoking and alcohol consumption. Results.\p=m-\Multiplelogistic regression analyses confirmed that case patients were more likely than control patients to be nulliparous (adjusted odds ratio [OR], 5.4; 95% confidence interval [CI], 2.8 to 10.3) and that preeclampsia in a previous pregnancy greatly increased the risk in a subsequent one (adjusted OR, 10.8; 95% CI, 1.2 to 29.1). However, regardless of parity, preeclamptic women were also more likely to be of high body mass (adjusted OR, 2.7; 95% CI, 1.2 to 6.2), to work during pregnancy (adjusted OR, 2.1; 95% CI, 1.1 to 4.4), and to have a
=

Objective.\p=m-\Todetermine, in a multivariate analysis, risk factors for preeclampsia that could be observed early in pregnancy and to establish whether these risk factors are different for nulliparas and multiparas. Design.\p=m-\Acase-control study of preeclampsia.

sally agreed on. Furthermore, although

family history of hypertension (adjusted OR, 1.7; 95% CI, 0.92 to 3.2). Having a previous history of a spontaneous abortion was protective but only in multiparous women (adjusted OR for multiparas, 0.09; 95% CI, 0.02 to 0.48). In contrast, being black was a significant risk for preeclampsia but only in nulliparous women (adjusted OR for nulliparas, 12.3; 95% CI, 1.6 to 100.8). Conclusions.\p=m-\Thereare a number of risk factors for preeclampsia that may be determined early in a woman's pregnancy. Multiparas and nulliparas share certain risk factors but not others. A cohort investigation is needed to determine the ability of these risk factors to predict who develops preeclampsia.
PREECLAMPSIA, a hypertensive dis order of pregnancy, is a major contribu tor to maternal mortality, premature birth, intrauterine growth retardation,
From the Programs of Maternal and Child Health and Epidemiology and The Northern California Occupational Health Center, School of Public Health, University of California, Berkeley (Dr Eskenazi); Reproductive Epidemiology Program, California Department of Health Services, Berkeley (Dr Fenster); and Division of Re-

(JAMA. 1991;266:237-241)

and perinatal mortality. Low-dose aspi rin has been found to be effective in preventing its development.1"1 Howev er, the use of aspirin during pregnancy may not be without risk.5 Hence, it is
See also
260.

search, Kaiser Permanente Medical Center, Oakland, Calif (Dr Sidney). Reprint requests to 312 Earl Warren Hall, School of Public Health, University of California, Berkeley, CA
94720 (Dr Eskenazi).

important to identify women who are likely to be at high risk for developing preeclampsia and to identify them early in pregnancy so that they can benefit
from intervention. The
causes

of and

Downloaded From: http://jama.jamanetwork.com/ on 05/16/2012

The relationship of body habitus and preeclampsia has been debated for al
most 200 years. Some authors have con

Table 1.Criteria for Inclusion In the Case and in the Control

I Case

Groups
a

protection against preeclampsia in a subsequent pregnancy.23 The present case-control study exam ines the interrelationship of these po
tential risk factors as well as others in a multivariate statistical model. We focus on those potential risk factors that can be determined early in pregnancy be fore the usual time of diagnosis (after 28 weeks' gestation). We also determine whether these risk factors for pre eclampsia differ in multiparous and in

cluded that frail women are more sus ceptible,17 whereas others claimed heavyset women are more vulnera ble.1819 Again, Chesley7 suggests that the association with heavier weight is an artifact of including women with essen tial hypertension in studies of pre eclampsia. Nevertheless, the WHO Study Group20 concludes that high prepregnancy body mass increases a wom an's risk for preeclampsia. Some attributes seem to protect women from developing preeclampsia. For example, cigarette smoking has been consistently demonstrated to pro tect women against preeclampsia.21,22 Its effect is dose related and independent of body mass.22 Also, a previously termi nated pregnancy, especially if it ends later in gestation, may provide some

A. Blood pressure (at least 2 readings fulfilling criteria after 20 wks' gestation*) 1 With a recorded baseline blood pressure (before 20 wks' gestation): from baseline,
mm

Group

reading of 22 + ; or 2. 2300 mg of protein in a 24-h urine collection C No history of diabetes or chronic hypertension II. Control Group A. Blood pressure 1. With a baseline blood pressure: a change from baseline In MAP <20 mm Hg in any reading*; or 2. Without baseline blood pressure: all blood pressures after 20 wks <MAP of 105 mm Hg B. Urinary protein 1. All dipstick measurements* <1 + after 20 wks; and 2. <300 mg of protein in all 24-h urine collections C No medical history of diabetes or chronic hypertension
1. Two
+ or 1

B.

2. Without baseline blood pressure: MAP >105

Hg; or

change in MAPt

220

mm

Urinary protein (after 20 wks' gestation) urinary protein dipstick measurements} of 1

Hg

'Including during labor. tMAP indicates mean arterial pressure. MAP [systolic
=

JNo more than one of these can occur during labor.

(diastolic)]/3.

case group on the year of delivery and the Kaiser Permanente hospital where the delivery occurred. Controls were also required to meet the blood pres sure, urinary protein, and medical his tory criteria listed in Table 1. Of the 260 eligible controls, 93 failed to meet the blood pressure criteria, 34 others failed to meet the urinary protein criteria, and one additional woman had a history of chronic hypertension, leaving a total of 132 controls (52 nulliparas, 80

multparas).

nulliparous women.
METHODS

Participants
Cases and controls were drawn from who delivered live births and stillbirths in 1984 and 1985 at the 10 hospitals that constitute Kaiser Per manente of Northern California. Case Selection.Cases were select ed from the 263 women who received a discharge diagnosis of severe pre
women

Procedure Information was abstracted from medical charts by a trained abstractor using a standard form. The abstractor was blinded to our hypotheses, the dis charge diagnoses, and the criteria for cases and controls. Information ab stracted included such sociodemographic characteristics as maternal age, marital status, race, religion, and the woman and her partner's occupation; past and present medical and pregnancy history (diabetes, chronic hyperten heart disease, seizures), including per tinent family medical history; life-style habits (alcohol and tobacco); height and prepregnancy weight; and all measured blood pressures, urinary protein con centrations, and weights during preg nancy. Information was obtained pri marily from the prenatal record forms, which are similar across the Kaiser sys tem. However, if certain information was not available in the pregnancy rec ord, the medical chart for a defined time period preceding the pregnancy was searched. This information included height (recorded at any time after the age of 18 years), prepregnancy weight (up to 3 months preceding the pregnan cy), race (any time), history of toxemia (any pregnancy record prior to the in dex pregnancy), family history of hy pertension (up to a year prior to the pregnancy), and marital status (up to 3 months prior to the pregnancy).

sion, codes 642.5 and 642.6). In addi tion, cases had to meet the criteria for blood pressure, urinary protein, and medical history shown in Table 1. Of the 263 eligible case patients who received
did not meet the blood pressure criteria, another 34 did not meet the urinary pro tein criteria, and 14 others had a history of either diabetes or chronic hyperten sion. A total of 139 cases remained (105 nulliparas, 34 multparas). Control Selection.Potential con trols (n 260) were randomly selected from the discharge data tape from those women who had not received a dis charge diagnosis of any hypertensive disorder of pregnancy (International Classification of Diseases, Ninth Revi sion, codes 642.0 to 642.9). The control group was frequency matched to the
=

eclampsia or eclampsia (International Classification of Diseases, Ninth Revi

sion, toxemia, urinary

tract

infection,

discharge diagnoses of preeclampsia, 76

Job information was obtained from the prenatal history form usually com pleted by the clinician at the first prena tal visit; information about the length of employment during pregnancy was not consistently provided on the medical chart and therefore was not abstracted. Job title and industry were coded using the Bureau of the Census codes.24 Social class (working class and non-working class or professional/manager) was esti mated from the occupation of the wom an and from her partner based on a scheme developed by Krieger.26 For comparison of cases and controls on categorical variables, crude odds ra tios (ORs) and 95% confidence intervals (CIs) were calculated. Those variables that were debated in the literature as important risk factors for preeclampsia or that produced a point estimate at a value <.10 on the univariate analysis were entered into both a stepwise and a backward multiple logistic regression model. Variables that were selected (P<.10) by either the stepwise or the backward regression procedures were entered together as the main effects in the final model. In addition, all one-way interaction terms with parity were in cluded to determine if risk factors dif fered between nulliparas and multpa ras. Using a stepwise procedure, with main effects forced into the model, the significant interaction terms were de termined (P<.10). Confidence intervals for interaction terms were determined by the Wald method.26 The final model was rerun for those who had blood pres sure measurements before 20 weeks' gestation. Women who gave birth to twins were excluded from all multivari ate modeling. To include as many wom en as possible in the multivariate mod els, dummy variables for missing information were constructed for each of those variables with information missing on more than 35 women (ie, al cohol consumption, body mass, weight were per gain). All statistical analyses formed using SAS 5.18. "

Downloaded From: http://jama.jamanetwork.com/ on 05/16/2012

RESULTS

Table

2Sociodemographic Characteristics of Preeclamptic Cases and Controls

Cases

Sociodemographic Characteristics and Medical History Table 2 shows the demographic char acteristics of cases and controls. Case patients averaged 27 years of age and were, on average, a year younger than control patients (P =. 14). Case patients were more likely than control patients to be living apart from their partner, to be black, and to be employed. However, for women who worked, a similar pro portion in both groups had professional or managerial jobs (30% of cases vs 27% of controls). Similarly, 29% of the part ners of both case and control patients had professional or managerial jobs. Table 3 presents the pregnancy and medical history of cases and controls. As expected, case patients were almost five times more likely than control pa tients to be nulliparous and twice as likely to be primigravid. Among women who had been pregnant before, more than twice the proportion of case pa tients in comparison with control pa tients had a history of therapeutic abor tion, but half the proportion of case patients in comparison with control pa tients had a history of spontaneous abortion. Six of the case pregnancies ended in twin births (two for multpa ras, four for nulliparas), whereas none of the controls gave birth to twins. A larger proportion of case patients were of higher body mass and a smaller percentage were of lower body mass. Case patients weighed, on average,
2.7 kg more prior to pregnancy (62.7 kg vs 60.0 kg, P= .09). Smokers had lower body masses than nonsmokers and there were fewer smokers among the cases; however, even among nonsmok ers, case patients were less likely to be of low body mass (OR, 0.39; 95% CI, 0.11 to 1.20) and more likely to be of high

Controls
No. 13.7 8.6 113 19 82 18
10 7.6

Characteristic

No.
19 108

Crude Odds Ratio

1.83 0.64 1.00 2.08

95% Confidence Interval

0.81-4.12 0.29-1.40

Age, y

<20

-35 Marital status Married or living with Not living as married Race White

13.6
85.6

36 74

25.9 53.6 13.0 15.9 1.5 15.9 20.8

79.2 69.9 30.1

14.4 62.5 16.0 13.7 1.0

6.9

1.12-3.85

Hispanic
Asian Other Black

18 22
22

1.00

2.57
1.00

1.14-5.82

Employed during
No Yes Woman's

pregnancy
27 103 72

47

36.4
63.6

82 60
22

1.25-3.85

occupation Working class

Non-working class

73.2 26.8

0.85 1.00

0.45-1.62

Table 3.

Pregnancy and Medical History of Preeclamptic Cases and Controls*

Cases
No.
%

Controls No.

Crude

Odds Ratio

95% Confidence Interval

Parity
0

Previous Pregnancy History

105

34

75.5 24.5 43.2

56.8

52
80

39.4

60.6 25.8
74.2

4.75 1.00 2.19

1 00

282-8.00

Gravidity
1 Previous
No

60 79

history of SABt

63

79.8
20.2

65.3 34
62

Yes

0.48 63.3
2.16

0.24-0.95

Previous history of TABt No Yes

Week at 1 st s13

35
44

443 55.7

36

1.18-3.96

prenatal visit

Present
102
37

Pregnancy
91 41 132
68.9

73.4

26.6 95.7
4.3 4.3

0.81 Limited

0.47-1.36

No. of infants

bodymass(OR, 2.1;95%CI, 1.0to4.5). Although the pregnancies of case pa tients ended approximately 4 weeks earlier than those of control patients
(mean, 35.1
vs

Singleton
Twins

133

dispersion

Prepregnancy Quetelet index, kg/m2 High (>25.8) Weight gain by week 20, kg
<5.34

Low (<18.9)_ Mid (-18.9, S25.8)

78

associated with preeclampsia. Approximately 10% of the case pa tients smoked during pregnancy, and case patients were less likely to smoke than control patients. Similarly, case patients were less likely to drink alco hol, with only about 20% consuming al cohol during pregnancy. About 65% of case patients and 48% of the control group neither drank alcohol nor smoked. The case and control groups were comparable in their history of heart dis ease, renal disease, and urinary tract

pa tients gained more weight on average during the course of their pregnancy (15.8 vs 14.3 kg, P .09). This result is expected given the increase in edema
=

39.2

weeks),

case

32 46
73

67.8 27.8
38.7

75 16 55

70.8

1.00 1.92
1.00

0.97-3.85

49.6
50.4

-5.34

56 Habits

0.92-2.63

Smoked cigarettes No Yes Drank alcohol No Yes

Previous No Yes No

123 13

90.4

103

82.4
17.6

1.00

0.50 1.00 0.55

0.24-1.03

79.7 25
20.3

75

69.2

35

31.8

0.30-0.99

history of toxemia}

Medical
29 8

History
74

77.4

96.1 2.3

5.8 56.5 87
39

1.00 5.95 1.00 1.72

1 72-30.03

Family history of hypertension

Yes 57

43.5

69.0 31.0

1.03-2.87

jonly

*SAB indicates spontaneous abortion; and TAB, therapeutic abortion. fOnly includes those women who had been pregnant previously, includes those women who had given birth previously.

Downloaded From: http://jama.jamanetwork.com/ on 05/16/2012

infection. Case patients were over five times (albeit nonsignificantly) more likely than control patients to have a history of seizures prior to pregnancy (OR, 5.6; 95% CI, 0.6to 49.1). However, none of the 11 women diagnosed as eclamptic had a previous history of sei zures. Preeclampsia in a previous preg nancy produced a sixfold increased risk. Two control patients and one case pa tient had gestational diabetes during the present pregnancy. For the most part, case patients did not differ from control patients in their family medical history; approximately equal proportions of cases as controls had a positive family history for diabe tes, renal disease, heart disease, and seizures. One noteworthy exception is that more case patients than control pa tients had family members with a histo ry of chronic hypertension (OR, 1.7).
Multivariate Analysis Variables (as categorized in Tables 2 and 3) entered into the multiple regres sion models included race, employment status, maternal age, parity, prepreg nancy body mass as measured by the Quetelet index (weight/height2), weight gain in the first 20 weeks of gestation, smoking during pregnancy, alcohol con sumption during pregnancy, history of toxemia in a previous pregnancy, family history of hypertension, year of deliv ery (1984 or 1985), marital status, week of gestation at first visit, history ofther apeutic abortion, and history of sponta neous abortion. Because of missing data, the multivariate models contained a sample of 235 women (115 cases and 120 controls). As shown in Table 4, being nulliparous (OR, 5.4), working (OR, 2.1), hav ing a family history of hypertension (OR, 1.7), and having high prepreg
nancy
common

Table 4Multivariate

Logistic Regression Model for Predicting Case-Control Status (n 235)*

=

Intervalf Nulliparous (1 ) vs Multiparous (0)_5J4_2.8-10.3

Black

Variable

Odds Ratio

Adjusted

95% Confidence

Body mass Low body mass (1) vs Medium (0)_043_0.13-1.4 High body mass (1) vs Medium (0)_2/7_1.2-6.2 Smoked (1)vs Not (0)_045_0.18-1.1 Worked during pregnancy (1) vs Unemployed (0) 2.1 1.1-4.4 1.7 0.92-3.2 Family history of hypertension (1 ) vs None (0) 0.31 0.13-0.74 History of spontaneous abortion (1) vs No history 10.8 1.2-29.1 History of preeclampsia (1 ) vs No history (0)
*Twin pregnancies were excluded from model; because all twins occurred in the case group, there was limited dispersion when twin pregnancy was included. fAII variables listed met the P<. 10 criteria for inclusion in the model with the exception of low body mass (P= .16)
which
was

(1)vsNonblack(0)_2j>_0.97-6.4

forced in.

Table 5Variables

Predicting Case-Control Status

That Differ in

Nulliparous

and

Multiparous Women
95% Confidence

(n 235)*
=

Variable Race Black Black

Odds Ratio* 12.3 0.68

Adjusted

Interval
1.6-100.8
0.13-3.5

nulliparas vs nonblack nulliparas multparas vs nonblack multparas

History of SAB Nulliparas with SAB vs nulliparas without SAB_089_0.20-3.9 0.09 0.02-0 48 Multparas with SAB vs multparas without SAB
^Multiple logistic regression model controlling for the main effects of parity, race, body during pregnancy, maternal history of hypertension, history of spontaneous abortion, preeclampsia. SAB indicates spontaneous abortion.
mass,

and

smoking, working previous history of

tective

(OR, 0.09). High body mass, family history of hypertension, and working during pregnancy posed a risk, and smoking was protective in both multiparous and nulliparous women.
COMMENT As expected, we found that being nul liparous and having a previous history of preeclampsia greatly increased a woman's risk for preeclampsia. Howev er, in regard to other risk factors that continue to be debated, we found that case patients were more likely to be heavier prior to pregnancy, a conclusion also drawn by the WHO Study Group,20 and, at least for nulliparous women, case patients were more likely to be black. Chesley7 wrote that the increased susceptibility for preeclampsia of black women and of heavyset women, as not ed by some authors, was doubtful and, if found, was probably due to including women with essential hypertension. It is unlikely that the increased risk associ ated with being black in nulliparous women and with high body mass found in the present study can be explained by essential hypertension. First, all wom en with a history of essential hyperten sion were excluded, and we have includ ed only women with evidence of both proteinuria and hypertension. Second, in a separate analysis, we repeated the final regression model presented in Ta ble 4, including only women who had a

clear increase in blood pressure during the course of pregnancy (ie, had a base line blood pressure measurement be fore 20 weeks); we found that the risk for high body mass (OR, 3.4) and for being black in nulliparous women (OR, 11.8) remained strong. Working during pregnancy more than doubled the risk for preeclampsia. This result, however, does not appear to be related to socioeconomic status. For example, employment in either a work increased women's risk relative to un employed women. Similarly, there was no difference in the proportion of part ners of case patients and control pa tients who held working-class and professional/managerial jobs. Howev er, although women from a wide range of social classes use the Kaiser Perman ente Medical Care Program for their health care, the extremes of income are

illustrate the dose-response relation ship of case status with increased body mass; low body mass is associated with a slightly reduced risk, although not sig nificantly, for preeclampsia (OR, 0.43). Smoking tended to have a protective effect (OR, 0.45). The interaction of body mass and smoking proved nonsig nificant. Also, among women who had been pregnant before, a diagnosis of preeclampsia in a previous pregnancy greatly increased the risk for being a

body

cases than in controls. Low mass was forced into this model to

body mass (OR, 2.7)

in

ing-class or professional/managerial job

were more

patient (OR, 10.8). Only two risk factors were found to operate differently in nulliparous and multiparous women (Table 5). Among nulliparas but not multparas, being black posed a substantial increased risk (OR, 12.3), and among multparas but not nulliparas, having had a previous spontaneous abortion was highly procase

probably underrepresented.28 Working during pregnancy has been implicated as a risk factor for pre eclampsia in a variety of studies. For example, a similar finding was observed in a cohort study of women exposed to organic solvents.29 In that study, wom en who worked longer into their preg

nancy (with solvents) were most at risk. Klebanoff et al30 reported that the risk of preeclampsia was higher in a group of female residents (100% employed) than in a control group consisting of spouses of male residents, 79% of whom worked

Downloaded From: http://jama.jamanetwork.com/ on 05/16/2012

during pregnancy. A study by Marcoux

sumably have less) protected against preeclampsia. Women who work may
et al31 found that increased leisure-time activity (of which working women pre

be more stressed and have different lev els of physical activity than non working women; however, further investigation is needed to determine the exact param eters of work that might increase risk. Cigarette smoking tended to protect against preeclampsia in this investiga tion. We also found that, on univariate analyses, alcohol consumption was pro tective. The apparent protective effect of alcohol may be associated with smok ing since it was not found to be signifi cant in the multivariate models. In stud ies of nonpregnant adults, cigarette smokers and light drinkers, particularly women, have been found to have lower blood pressures than abstainers.32,38 Al though tobacco and alcohol consumption may tend to protect against preeclamp sia, the well-known risk of both of these agents to the fetus outweighs the slight benefit. Moreover, Duffus and MacGillivray34 found that the perinatal death rate due to preeclampsia was higher among infants of smokers than of nonsmokers. We did not demonstrate that age, ei ther old or young, is a risk factor for preeclampsia. The slight increased risk of women under 21 years old observed on univariate analyses disappeared when parity was controlled for, as Chesley suggested. The increased risk at old er age was not evident in this sample, perhaps because women with essential hypertension were excluded. In addi tion, Lehmann16 did not show a clear increase in the incidence of preeclamp sia until after women reached the age of 45 years. We did not have any patients above 41 years of age. Having a previous history of sponta neous abortion protected against pre eclampsia in multiparous women. How ever, it does not appear that any previ ous pregnancy provides protection, in that having a previous therapeutic abortion did not protect against pre that previous pregnancy, especially one that ended later in gestation, provided protection against preeclampsia. In a recent study of Kaiser patients, clinical ly recognized spontaneous abortions oc curred later in gestation, on average, than therapeutic abortions.36 Except for being black and having a history of spontaneous abortion, the re gression models predicting risk of pre eclampsia were identical for nulliparous and multiparous women. The differen tial risk of these variables in the two groups may be related to the stringent criteria we used to define preeclamptic cases and controls. For example, when

applied the final multivariate model original sample of women who received a discharge diagnosis of severe preeclampsia or eclampsia (n 263) and the random sample of controls (n 260), we found that being black posed a signif icant increased risk in both nulliparous and multiparous women. Similarly, spontaneous abortion was significantly protective in both nulliparous and mul tiparous women. This finding suggests that a history of spontaneous abortion may be protective and being black may be a risk in both nulliparous and multi
we

to the

parous women with less severe forms of preeclampsia than that used to define inclusion in this study.

The usefulness of low-dose aspirin therapy and other preventive measures depends on the clinician's ability to iden tify women who are at high risk for preeclampsia early in their pregnancy. Therefore, we included in our logistic model only those potential risk factors that could be known by the 20th week of gestation. These risk factors should be examined in a cohort study to determine the model's ability to predict who devel ops preeclampsia. Risk factors differ somewhat for nulliparas and multpa ras; therefore, future investigations should include multparas in studies of preeclampsia but examine their risks separately from those of primparas.
The research was funded by a grant from the Kaiser Foundation Research Institute. We gratefully acknowledge the assistance of Bet ty Wong and William Frank in abstracting and coding the data. We also acknowledge Linda Grand, MPH, Marianne Sadler, MPH, Paul En glish, MPH, Irene Tekawa, MPH, and Robert Sholtz, MPH, for statistical consultation and ana lyses and Robert Hosang, MD, MPH, Gary Stew art, MD, Irva Hertz-Picciotto, PhD, Barbara Abrams, DrPH, and James Roberts, MD, for their review of and comments on the research. We also thank Debra Patterson for her editorial comments. References
1. Beauphils M, Donsimoni R, Uzan S, Colau JC. Prevention of pre-eclampsia by early antiplatelet therapy. Lancet. 1985;1:840-842. 2. Wallenburg HCS, Markovitz JW, Dekker GA, Rotmans P. Low-dose aspirin prevents pregnancy\x=req-\ induced hypertension and pre-eclampsia in angiotensin-sensitive primigravidae. Lancet. 1986;1:1-3. 3. Schiff E, Peleg E, Goldenberg M, et al. The use of aspirin to prevent pregnancy-induced hypertension and lower the ratio of thromoboxane A2 to prostacyclin in relatively high-risk pregnancies. N Engl J Med. 1989;321:351-356. 4. Benigni A, Gregorini G, Frusca T, et al. Effect of low-dose aspirin on fetal and maternal generation of thromboxane by platelets in women at risk for pregnancy-induced hypertension. N Engl J Med.

1986;155:1011-1016. Chesley LC, Annitto JE, Cosgrove RA. The remote prognosis of eclamptic women: sixth periodic report. Am J Obstet Gynecol. 1976;124:446-456. 10. Gleicher N, Boler LR, Norusis M, Del Grando A. Hypertensive diseases of pregnancy and parity. Am J Obstet Gynecol. 1986;154:1044-1049. 11. Sibai BM. Pitfalls in diagnosis and management of preeclampsia. Am J Obstet Gynecol. 1988;159:1-5. 12. Fitzgibbon G. The relationship of eclampsia to the other toxemias of pregnancy. J Obstet Gynaecol Br Commonw. 1922;29:402-415. 13. Davies AM, Czaczkes JW, Sadovsky E, Prywes R, Weiskopf P, Sterk VV. Toxemia of pregnancy in Jerusalem, I: epidemiological studies of a total community. Isr J Med Sci. 1970;6:253-266. 14. Nelson TR. A clinical study of preeclampsia. J Obstet Gynaecol Br Commonw. 1955;62:48-66. 15. Niswander KR, Gordon M. The Collaborative Perinatal Study of the National Institute of Neurological Diseases and Stroke: The Women and Their Pregnancies. Philadelphia, Pa: WB Saunders Co;
9.

preeclampsia-eclampsia in young primigravid women: subsequent pregnancy outcome and remote prognosis. Am J Obstet Gynecol.

Eklampsien i Danmark i Aarene Copenhagen, Denmark: Busck; 1933. 17. Denman T. An Introduction to the Practice of Midwifery. New York, NY: James Oram; 1821.
1918-1927.

1972. 16. Lehmann K.

18. Baird D. Social factors in obstetrics. Lancet.

19. Barron SL. The epidemiology of human pregnancy. Proc R Soc Med. 1968;61:120-124. 20. The Hypertensive Disorders of Pregnancy: Report of a WHO Study Group. Geneva, Switzerland: World Health Organization; 1987. World Health Organization Technical Report Series 758. 21. Palmgren B, Wahlen T, Wallander B. Toxemia and cigarette smoking during pregnancy. Acta Obstet Gynecol Scand. 1973;52:183-185. 22. Marcoux S, Brisson J, Fabia J. The effect of cigarette smoking on the risk of preeclampsia and gestational hypertension. Am J Epidemiol. 23. MacGillivray I. Hypertension in pregnancy and its consequences. J Obstet Gynaecol Br Commonw.

1949;1:1070-1074.

1989;130:950-957.

24. 1980 Census

1961;68:557-569.

eclampsia. MacGillivray23 suggested

5. Hertz-Piccioto I, Hopenhayn-Rich C, Golub M, Hooper K. The risks and benefits of taking aspirin during pregnancy. Epidemiol Rev. 1990;12:108\x=req-\ 148. 6. Roberts JM. Pregnancy-related hypertension. In: Creasy RK, Resnick R, eds. Maternal-Fetal Medicine. Philadelphia, Pa: WB Saunders Co; 1984:703-752. 7. Chesley LC. History and epidemiology of preeclampsia-eclampsia. Clin Obstet Gynecol. 8. Sibai BM, El-Nazer A, Gonzalez-Ruiz A. Severe

1989;321:357-362.

dex of Industries and Occupations. Washington, DC: US Bureau of the Census; 1980. 25. Krieger N. Women and social class. J Epidemiol Community Health. 1991;45:35-42. 26. Chiang CL. An Introduction to Stochastic Processes and Their Applications. New York, NY: Robert E Krieger Publishing Co; 1980. 27. SAS User's Guide, Version 5. Cary, NC: SAS Institute; 1985. 28. Hiatt RA, Friedman GD. The frequency of kidney and urinary tract diseases in a defined population. Kidney Int. 1982;22:63-68. 29. Eskenazi B, Bracken MB, Holford T, Grady J. Exposure to organic solvents and hypertensive disorders of pregnancy. Am J Ind Med. 1988;14:177\x=req-\ 188. 30. Klebanoff M, Shiono PH, Rhoads GG. Outcome of pregnancy in a national sample of resident physicians. N Engl J Med. 1990;323:1040-1045. 31. Marcoux S, Brisson J, Fabia J. The effect of leisure time physical activity on the risk of preeclampsia and gestational hypertension. J Epidemiol Community Health. 1989;43:147-152. 32. Klatsky AL, Friedman GD, Siegelaub AB. Alcohol and hypertension. Compr Ther. 1978;

of Population: Alphabetical

In-

4(12):60-68.

33. Friedman GD, Klatsky AL, Siegelaub AB. Alcohol, tobacco, and hypertension. In: Perspectives in Hypertension 1982. Hypertension. 1982;4(suppl

III):143-150.

1984;27:801-820.

34. Duffus GM, MacGillivray I. The incidence of preeclamptic toxaemia in smokers and nonsmokers. Lancet. 1968;1:994-995. 35. Goldhaber MK, Fireman BH, The fetal life table revisited: spontaneous abortion rates in three Kaiser Permanente cohorts. Epidemiology.

1991;2:33-39.

Downloaded From: http://jama.jamanetwork.com/ on 05/16/2012