EPIDEMIOLOGIC REVIEWS

Copyright © 1990 by The Johns Hopkins University School of Hygiene and Public Health Allrightsreserved

Vol. 12, 1990 Printed in U.S.A.

THE RISKS AND BENEFITS OF TAKING ASPIRIN DURING PREGNANCY IRVA HERTZ-PICCIOTTO,1 CLAUDIA HOPENHAYN-RICH,2 MARI GOLUB,3 AND KIM HOOPER4

Aspirin has been examined in relation to a wide range of adverse pregnancy-related outcomes: maternal and infant bleeding, low birth weight, prolonged labor, prolonged gestation, perinatal mortality, premature closure of the ductus arteriosus, cognitive deficits, congenital malformations, and childhood cancer. While the evidence for certain end points is not persuasive, some of the findings appear to be strong, consistent, and biologically plausible. For some outcomes, positive and null studies need to be reconciled; for others, the clinical or social significance assessed. Frequently, further study is required. Curiously, one of the most common adverse pregnancy outcomes, spontaneous abortion, appears not to have been examined in relation to aspirin use. In addition to the literature on risks, a number of potential benefits of aspirin specific to pregnancy have been studied. These include reductions in pregnancy-induced hypertension, preeclampsia, umbilical placental insufficiency, and their sequelae of low birth weight, premature labor, and still-

births; and repeat second and third trimester fetal losses resulting from the presence of lupus anticoagulant. Aspirin use during pregnancy is common (table 1) (1-10). When asked prospectively, more than 50 percent of women reported using aspirin during pregnancy (1). In a survey in Great Britain during the 1960s, aspirin was consumed for a mean of 31 days during pregnancy, and taking into account both the proportion of women taking the drug and the duration of consumption, aspirin was the most commonly used single substance after vitamins and iron (4). Salicylate was detected in 10 percent of the cord blood from 272 consecutive live births at a medical center in Alabama (11). Although aspirin use in the population has declined (12), consumption during pregnancy could increase as its efficacy in reducing toxemia (13-16) becomes more established. Because it is a common ingredient in other medications, some studies may have underestimated its use. That the consequences of such common exposures for the developing fetus are not well understood is, therefore, surprising. This paper presents a comprehensive and

Received for publication December 20,1989, and in final form July 2, 1990. Abbreviations: RR, risk ratio; IQ, intelligence quotient. 'Department of Epidemiology, University of North Carolina, Chapel Hill, NC; formerly, Department of Biomedical and Environmental Health Sciences, University of California, Berkeley, CA. ^Department of Biomedical and Environmental Health Sciences, University of California, Berkeley, CA. 'California Department of Health Services, Reproductive and Cancer Hazard Assessment Section, Sacramento, CA. California Department of Health Services, Repro-

ductive and Cancer Hazard Assessment Section, Berkeley, CA. Reprint requests to Dr. Irva Hertz-Picciotto, CB # 7400, McGavran-Greenberg Hall, University of North Carolina, Chapel Hill, NC 27599-7400. This work was supported by the University of California Health Effects Component of the Toxic Substances Research and Training Program, the National Institute of Environmental Health Sciences, and the California Department of Health Services. The authors thank those who made helpful comments on earlier drafts of this paper: Dr. Allen Wilcox, Dr. James Huff, Dr. David Savitz, Dr. Bernard Schwetz, Kim Waller, Paige Hornsby, Beth Whelan, Dr. Joe Haseman, and Dr. Andrew Rowland.

OVERVIEW

108

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

109

ASPIRIN USE IN PREGNANCY TABLE 1

Prevalence of aspirin use in pregnancy Location

Time period of pregnancies

Authors (reference)

No.

Twelve US cities

1959-1965 Heinonen et al. (1)

South Wales

1964-1966 Richards (2)

833

Scotland

1960s

911

Nelson and Forfar (3) Forfar and Nelson (4)

Rochester, NY 1960s Bleyer et al. (5) Brisbane, Australia 1971-1972 Finnigan et al. (6)

50,282

67 250

Percent using aspirin in pregnancy

29.6 64.0 14.4 24.2 54

69

12.8

Australia

early 1970s Collins and Turner

-2,000

(7)

Massachusetts

1973-1975 Rothman et al. (8)

Seattle, WA Michigan (Medicaid recipients)

1974-1975 Streissguth et al. (9) 1981-1983 Piper et al. (10)

1,254 1,529 18,886

6.6 2.9 16 46 2.9 1.5 1.1

critical review of the published literature on the risks and benefits of aspirin use during pregnancy. Several published reports have summarized the findings of studies on aspirin use in pregnancy (1721), and each concluded that aspirin should be avoided, at least in the last trimester. These reviews are generally more limited in scope than the present paper, some are out of date, and most have a clinical rather than epidemiologic perspective. The present paper takes an in-depth look at how the quality of studies bears on the interpretation of findings. Unpublished papers were not reviewed, but for the epidemiologic literature, the intention was to be comprehensive, especially with respect to the adverse effects. For animal studies, our review did not include 1) studies in which aspirin was given in the diet rather than in discrete doses, as it is in humans, 2) most studies of aspirin interactions with other teratogens, 3) stud-

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

Notes on exposure

First 4 months All pregnancy 1st trimester only 2nd and 3rd trimesters 5.8% were corroborated by physician, hospital, or prescription Last trimester only No data on "ever" use Self reported daily use of two or more doses of analgesic No data on "ever" use Regularly Daily Around time pregnancy began First half of pregnancy No data on "ever" use 1st trimester purchases only 2nd trimester purchases only 3rd trimester purchases only Underestimates if aspirin purchased at a prior date was consumed during pregnancy

ies with other salicylates but not aspirin, and 4) studies where aspirin was used as a positive control. Some of the data on risks and benefits appear to be contradictory. Low birth weight is reported to be both increased and reduced by aspirin consumption. More intriguingly, the same effect may be both beneficial and detrimental, depending on the particular pregnancy or on the gestational age. Prolonging a pregnancy that is ending prematurely is likely to improve the outcome, whereas prolonging a postmature pregnancy is likely to be harmful. The same is true for closure of the ductus arteriosus postterm versus preterm. Ultimately, the key to making sense of these data lies in understanding the roles of dose, timing, and individual differences in the physiology of pregnancy. Before discussing the data for each end point, we briefly review first, the pharmacology and toxicology of aspirin and sali-

110

HERTZ-PICCIOTTO ET AL.

cylic acid, and second, the salient epidemiologic issues relevant to the literature on aspirin in pregnancy. The animal and human evidence on each adverse effect is then described and summarized. Where relative risks, confidence intervals, or significance levels were not reported, they were calculated for this paper. The penultimate section reviews the benefits associated with low-dose aspirin use in humans for certain high-risk pregnancies. The paper concludes with recommendations for further research. PHARMACOLOGY AND TOXICOLOGY

Aspirin, or acetylsalicylic acid, is a member of a group of pharmacologic agents for which the compound salicylate is the common active agent. The major pharmacologic effects of salicylates include antiinflammatory response, inhibition of platelet aggregation, analgesic action, and antipyretic action. The first two effects are thought to be mediated primarily by inhibition of prostaglandin synthesis while the mechanisms for the last two are not well understood (22). Aspirin acts on prostaglandin synthesis chiefly by inhibition of cyclooxygenase, a rate limiting enzyme in the conversion of arachidonic acid to prostaglandins (23). A single aspirin is 325 mg and an extrastrength tablet is 500 mg. Thus, typical analgesic or antipyretic doses, e.g., two tablets, taken two to four times a day, is 1,3004,000 mg/day. Antiinflammatory therapy may involve 3,000-5,000 mg/day, while very low doses currently being used to prevent preeclampsia are under 150 mg/day, i.e., an order of magnitude lower than the doses used for pain or fever. Both aspirin and its metabolite salicylate are biologically active. Aspirin is rapidly converted to salicylate in the circulation (half-life of 15-20 minutes) whereas salicylate is more gradually eliminated (half-life of 2-3 hours at low doses, 12 hours at therapeutic antiinflammatory doses, longer at higher doses) (23). Toxicologic effects, however, may persist beyond the period of elevated plasma levels. Because of satura-

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

tion of salicyluric acid formation and one glucuronide conjugation pathway (23), the elimination kinetics of aspirin are dosedependent. Thus, both extent and duration of elevated plasma salicylate levels will be considerably reduced with very low dose (less than 150 mg) aspirin therapies as compared with analgesic/antipyretic doses. The major toxic effects of salicylates are bleeding and ulceration of the gastric mucosa, liver and kidney damage, hearing loss, and with acute high dose poisoning, upset of acid-base balance with accompanying symptoms of metabolic acidosis, hyperventilation, and respiratory alkalosis (23). Death can occur with ingestion of 10-30 g (23). Adverse effects do occur at therapeutic doses. Fecal blood loss rises 10-fold with administration of typical antiinflammatory regimens of 4-5 g of aspirin per day (23). Bleeding times are doubled from a single dose of 650 mg (two regular strength tablets) and this effect will last for a period of 4-7 days (23). Salicylates have a common spectrum of activity in the variety of species tested. However, effective doses vary considerably among species partly because of differences in plasma protein binding affinity. Species with relatively low plasma protein binding affinity (30-40 percent in rats and dogs) are more sensitive to aspirin than those with higher protein binding affinity (60-70 percent in monkeys and humans) at the same albumin concentrations (24). Thus, the effective pharmacologic and toxicologic doses are not readily extrapolated from animal to human studies. Nevertheless, there is little reason to expect qualitatively different effects in different species. When administered to the mother, salicylates transfer readily to the fetus in animals (25) and humans (11, 26) where pharmacologic activity, such as inhibition of prostaglandin synthesis, also occurs in both animals (27) and humans (28). Because of its short half-life, only a small amount of unmetabolized aspirin reaches the human fetus (26). Compared with the adult, the fetus and newborn have lower plasma pro-

ASPIRIN USE IN PREGNANCY

tein binding of salicylates in animals (29) and in humans (30, 31), reduced metabolic activity (in particular, glucuronidation) in humans (32), and less efficient elimination, also in humans (32). All of these factors can lead to higher plasma concentrations. In human neonates whose mothers had taken salicylates before delivery, plasma concentrations of salicylate were as much as four times higher than in the mother, and half-lives of salicylate ranged from 4.5 to 11.5 hours compared with 3 hours in adults (32). Thus, salicylates taken during pregnancy are likely to have similar actions in the fetus as in the mother, but the actions may be greater in the fetus and newborn because of relatively higher plasma concentrations. ISSUES IN EVALUATING EPIDEMIOLOGIC STUDIES OF ASPIRIN AND PREGNANCY

The main issues pertinent to the evaluation of the epidemiologic studies are exposure ascertainment, confounding by conditions for which aspirin was used or by other drug exposures, the testing of multiple hypotheses, and publication bias. The quality of exposure data was by far the prime determinant of the strength of evidence in these studies. While studies varied greatly in how clearly exposure to aspirin was defined and how effectively its use ascertained, quantitative data on aspirin consumption late in pregnancy were generally available. In a few studies, even the timing of exposure was collected, to the day (33, 34). Unfortunately, high quality exposure information was not available for studies of aspirin use early in pregnancy. The timing of ascertainment, prospective or retrospective, is also relevant. Prospective ascertainment has the advantages of improving accuracy and limiting bias in women's reporting of exposure based on knowledge of pregnancy outcome. Such bias is more likely when the outcome is one of which the woman is aware and potentially concerned (e.g., congenital malformations, low birth weight), as opposed to outcomes which are measured by the prac-

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

111

titioner or of which the woman may be unaware (e.g., quantity of blood loss). However, with prospective ascertainment of the woman's exposure, awareness by the physician can lead to biased ascertainment of the outcome if aspirin use is believed to have an effect. Also, even within a prospective exposure assessment, studies vary in the frequency and timing of data collection, e.g., an interview once during pregnancy (9), an interview at each prenatal visit (1), or a daily diary (33). On the other hand, several studies relying on retrospective recall of exposure controlled for biased reporting by the use of "diseased" controls. The investigators compared the particular defects of interest with other birth defects (35, 36). Since reporting of aspirin would not be expected to differ by type of defect, bias resulting from differential misclassification was unlikely. Validation of self-reports by medical records is believed to improve data quality. In the case of aspirin, requiring such "validation" is likely to lead to severe underascertainment of exposure, since this widely used drug is mainly self administered. Investigators in Scotland (3, 4) were able to corroborate from records only about 10 percent of women's self-reports of aspirin consumption (see table 1). In a study where pregnant women kept daily diaries of the medications they took, 80 percent of the drugs were taken without the physician's knowledge (5). An exception where the use of medical or prescription records would be appropriate is in studies of those suffering from specific medical conditions (e.g., rheumatoid arthritis (37)) for which higher therapeutic doses of aspirin (~3,000 mg) are prescribed. When self-reports are used, the type of question asked (i.e., open-ended (8) vs. prompting by a list (34-36)) influences the extent of ascertainment, particularly for nonprescription drugs (38). Many studies using interviews do not indicate how the question was asked. Open-ended questions result in far lower reporting of aspirin consumption than questions which name prod-

112

HERTZ-PICCIOTTO ET AL.

ucts (11, 38) or indications for use (38). Whether open-ended questions introduce differential error (with the potential for bias away from the null) or just random error (causing bias toward the null) is not clear. Furthermore, some studies assessed exposure through a detailed analysis of all medications in which aspirin could be a constituent (e.g., 1, 34-36), while others appear not to have done so (8). Other factors external to the study can influence reporting of exposure. For instance, during the time period of the large prospective Collaborative Perinatal Project study of over 50,000 pregnancies (19591965), the teratogenicity of thalidomide was discovered and made public. Reports of total drug use during pregnancy were far higher in the latter part of the study period (1, pp. 264-5), suggesting differential reporting over time. Any analysis of data from this study should take such changes into account. A valuable addition to self-reports of aspirin consumption is measurement of internal dose (e.g., plasma salicylate levels) (7) or target tissue effects (e.g., platelet aggregation) (34), but these are not relevant for studies concerned with exposure much earlier in time (more than a few weeks previous). Two other germane issues are confounding from other drugs and confounding by the condition for which aspirin was prescribed. The former can occur because women who use aspirin may be more likely to use other drugs. Unfortunately, some studies which collected detailed drug exposure information were conducted before multivariate methods were widely available (e.g., 3), although simple stratification could have been performed. To control for confounding from the condition which prompted aspirin consumption, one group of investigators used a comparison group of women with the same type of conditions as the exposed group (rheumatoid arthritis), but who received no aspirin (37). Other researchers have included a group who used acetaminophen, since it is usually taken for

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

the same conditions as aspirin (9, 39). Because aspirin is used as an antipyretic, in studies that do not control for the underlying medical problem, fever or infection could be the harmful agent. Quite a few studies on congenital malformations were designed to look at a large number of exposures, and some also looked at a wide range of malformations. Thus, a number of false positive associations may occur by "chance." However, for any particular positive association, the probability of it being the result of random fluctuations is no higher than when only one test is conducted. The fact that other hypotheses were tested has no bearing on this probability. The "universe" of tests is never known, and when only a single hypothesis is tested, it is still only one in a larger universe of many studies. The goals of scientific inquiry are not consistent with invoking a "penalty" for looking at more information (15a). Moreover, to require a higher "significance level" is to increase the probability of not "detecting," and hence possibly ignoring, a true and meaningful association. Having a specific hypothesis can render the results of a study more credible when 1) prior knowledge (from previous research in the same or another species, or from known biologic mechanisms) brings plausibility or consistency and, therefore, greater weight in favor of a hypothesis, or 2) a clear focus on a specific exposure leads to more attention being paid to the methods for data collection. It is prior scientific information, not what is in the investigator's mind per se before collecting or analyzing the data, that affects the credibility of a finding. For these reasons, we have not given greater weight to associations from studies of a single hypothesis. Instead, quality of exposure data and control of biases and confounding are the critical considerations. Finally, as in any review, the problem of publication bias is a concern. Because no attempt was made to identify unpublished investigations, it is possible that the liter-

ASPIRIN USE IN PREGNANCY

ature is weighted with "positive" findings because of a lower likelihood that studies showing no effect of aspirin will be written up or selected for publication. Unfortunately, detection of such bias is difficult, and determining its magnitude is even more problematic. Certainly, the replication of positive findings is a prerequisite for ruling out chance as an explanation; replication with different study designs and study populations reduces the likelihood that bias or confounding is responsible for positive associations, and strengthens the case for causation. ADVERSE EFFECTS OF ASPIRIN CONSUMPTION LATE IN PREGNANCY

A number of end points considered in this review were studied in relation to aspirin use late in pregnancy: bleeding and other hemostatic abnormalities in both the mother and newborn, delays in the onset of labor, prolongation of labor, low birth weight, perinatal mortality, and premature closure of the ductus arteriosus. For each end point, the animal data are presented first, followed by the human data and a conclusion. Epidemiologic studies of these end points (listed in table 2) were all of the cohort design, though exposure ascertainment was prospective in some studies and retrospective in others. The strengths and weaknesses of studies that examined more than one end point are discussed in detail under the first end point reviewed. For this reason, the reader may need to refer to table 2 throughout this section for an overview of the study population, the exposure information, and the results. The first two subsections discuss hemostatic abnormalities. This term is used here to refer to abnormal bleeding tendencies as measured either clinically or by laboratory tests. Since aspirin is known to affect bleeding through its inhibition of platelet aggregation, numerous studies have investigated bleeding tendencies at the time of delivery among women and their offspring after exposure during pregnancy.

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

113

Maternal hemostatic abnormalities Animal data. A detrimental effect of aspirin on maternal hemostasis was demonstrated in one study on rats (40). Compared with controls, rats administered aspirin at a dose of 30 mg/kg twice a day late in gestation (days 18-21) had a greater frequency of excess maternal blood loss at delivery (5 of 9 vs. 0 of 10). This dose, equivalent (on a weight per body weight basis) to about 7-12 aspirin tablets per day in humans, was lower than the dose required for teratogenic effects, though the timing was different as well. Human data. Two epidemiologic studies provide strong evidence of aspirin-related increases in overt maternal bleeding (see table 2). The first, a retrospective study by Lewis and Schulman (37), was based on a review of obstetric records in New York during the period 1951-1971. Exposed women (n = 103) were those who suffered from rheumatoid arthritis or some other degenerative musculoskeletal condition and were prescribed 3,250 mg of aspirin (the equivalent of 10 regular aspirin tablets) or more daily, for at least the last 6 months of pregnancy. Two control groups were used: 1) 52 women with the same type of conditions, but who were not prescribed aspirin or any other prostaglandin inhibitor during their pregnancy, and 2) 50 women with no such conditions and who were not prescribed aspirin for any reason. The aspirin group had a 39-45 percent greater blood loss (an average of 100 ml) during delivery than did either of the control groups (p < 0.025). All babies were delivered vaginally. By using two control groups, this study clearly suggests that aspirin, and not the condition for which it was prescribed, was responsible for the excess bleeding. Furthermore, the observed effect may have been underestimated given that exposure information was limited to medical records, thus not accounting for either noncompliance in the treated group or self-administration of over-the-counter aspirin by the controls. While it could be argued that the severity of musculoskeletal

TABLE 2

Studies of effects of aspirin use late in pregnancy Study groups Study (reference)

wis and Schulman, 1973 (37)

Ascertainment of exposure

Outcomes*

(+) Maternal bleeding (+) Prolonged gestation (+) Prolonged labor (—) Low birth weight

Size and definition

Exposure to aspirin

(1) n = 103; aspirin used for rheumatoid arthritis or other musculoskeletal disease (2) n = 52; no aspirin, also had rheumatoid arthritis or other musculoskeletal disease (3) n = 50; no aspirin, no rheumatoid arthritis or other musculoskeletal disease

(1) 3,250 mg daily for at least last 6 months of pregnancy (2) None

Medical records only

(3) None

uart et al., 1982 (34)

(+) Maternal bleeding (+) Impaired maternal platelet function (+) Infant bleeding (+) Impaired infant platelet function

(1) n = 10; aspirin used within 5 days of delivery (2) n = 7; aspirin used 6-10 days prior to delivery (3) n = 34; no aspirin in 10 days prior to delivery

(1) 5,000-10,000 mg total, -1,500 mg/day (2) 5,000-15,000 mg total, -2,000 mg/day (3) None

llins and Turner, 1975 (7)

(+) Maternal bleeding (+) Prolonged gestation (—) Prolonged labor

(1) n = 63; daily use of aspirin/caffeine with salicylate or phenacetin throughout pregnancy (2) n = 81; use of same 1-6 times per week (3) n = 63; no use of salicylates throughout pregnancy

(1) 2-12 tablets or powders/day

rner and Collins, 1975 (74)

(+) Perinatal mortality (+) Low birth weight

Same as Collins and Turner above

Same as Collins and Turner above

Same as Collins and Turner above

rby and Schulman, 1971 (43)

(+) Impaired maternal platelet function (+) Impaired infant platelet function

(1) n = 10; aspirin, but no other drug, used in week before delivery (2) n = 18; no aspirin or other drug in week before delivery

No dose information

Maternal recall at time of delivery

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

(2) 2-12 tablets or powders/day (3) None

Maternal recall before delivery (question named 100 drugs containing aspirin) Platelet malondialdehyde assay Clinical observations recorded blindly Aspirin users identified by urine tests Dose groups based on maternal recall shortly after delivery Maternal and cord bloods were tested for salicylates

ai et al., 1989 (46)

mack et al., 1981 39)

(+) Impaired maternal platelet function (—) Impaired infant platelet function (—) Infant bleeding (—) Prolonged labor (—) Prolonged gestation (—) Low birth weight

(+) Infant bleeding (intracranial hemorrhage)

Normotensive primigravidae (n = 40) no other drugs during trial, no aspirin in preceding 4 weeks, aspirin began at ~37 weeks gestation (1) n = 10 (7 for neonatal platelet function) (2) n = 10 (5 for neonatal platelet function) (3) n = 10 (5 for neonatal platelet function) (4) n = 10 (4 for neonatal platelet function) All very low birth weight or very premature infants (< 1,500 g or < 34 weeks gestation): (1) n = 17; aspirin used within 1 week of delivery (2) n = 20; acetaminophen used within 1 week of delivery (3) n = 71; neither aspirin nor acetaminophen used during last week of pregnancy

Clinical trial; women and investigators blind with respect to dose group (1) Placebo (2) 20 mg/day (3) 60 mg/day (4) 80 mg/day

(1) Range: 1-5 tablets total (325-2,500 mg total) (2) None

Maternal recall 2 days after delivery, prior to diagnosis of infant's intracranial hemorrhage

(3) None

yer and Breckenidge, 1970 (33)

(+) Impaired infant platelet function

(1) n = 14; aspirin used in last week of pregnancy (2) n = 17; no aspirin used in last 3 weeks of pregnancy

(1) From 320 mg once to 1,300 mg daily (2) None

Maternal diary of all medications taken in last month of pregnancy recorded prospectively Laboratory tests conducted blindly

apiro et al., 1976 (75)

(—) Low birth weight (—) Perinatal mortality

(1) n = 1,515; heavy aspirin use

(1) 8 or more days per month for at least 6 lunar months (2) Any other aspirin use (3) None

Maternal recall at prenatal visits

(2) n = 24,866; other (3) n = 14,956; none

*(+),P0.05.

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

116

HERTZ-PICCIOTTO ET AL.

disease may have been related to the use of founding because of conditions such as feaspirin, the biologic plausibility that such ver. However, given the magnitude of the diseases are related to bleeding tendencies associations (19-fold for bleeding, 2.7-fold is far lower than the plausibility of aspirin for the fall in hemoglobin), confounding from febrile illness seems unlikely to exbeing causally related to bleeding. Stuart et al. (34) observed similar effects plain the aspirin effect. In addition, 1) the at lower doses in a prospective study of pregnancies were normal, 2) these results women with normal, full-term pregnancies confirm an earlier study that did control admitted for delivery. The unique aspect of for the conditions for which aspirin was this study was the collection of drug histo- used (37), 3) an established biologic mechries for each of the 10 days prior to delivery, anism via impaired platelet function can allowing exploration of the effect of timing explain the results, and 4) all clinical and of aspirin consumption on maternal and laboratory observations were conducted neonatal hemostasis (see table 2). The 10 blindly with respect to exposure of the subwomen who took any aspirin within 5 days jects. Self-reports of exposure in the prebefore delivery had, on average, a 2.7-fold vious 10 days were obtained by interview greater drop in pre- to postpartum hemo- at the time of delivery and were validated globin as compared with 34 women who by platelet assays sensitive to the timing of took no aspirin (14 vs. 5.2 percent, p < aspirin ingestion. All of these factors argue 0.001). They also experienced a signifi- for the validity of these findings. cantly higher incidence of pronounced Further evidence comes from a study bleeding, with a risk ratio (adjusted for conducted in Australia (see table 2) that small sample size (41)) of 19 (p < 0.0001). compared 63 daily users of "headache powThe seven women who took aspirin be- ders/tablets" with a matched group of tween 6 and 10 days before delivery showed women who took no aspirin in pregnancy neither effect. However, laboratory tests of (7). Compared with the unexposed women, this latter group did show a detrimental daily users had significantly increased ineffect on platelet aggregation and malon- cidences of pre- and postpartum hemordialdehyde levels as compared with non- rhage (14 vs. 4 percent and 12 vs. 2 percent, users of aspirin. The doses taken in this respectively), anemia during pregnancy (41 study were about three to five aspirin daily, vs. 20 percent), and transfusions at delivery less than half the doses taken by the women (12 vs. 0 percent). Postpartum hemorrhage being treated for arthritis in the study by was defined as > 600 ml in the first 24 Lewis and Schulman (37). hours after delivery. Anemia was defined Although the study groups were small, as hemoglobin less than or equal to 10.5 g/ the findings of Stuart et al. (34) are consis- 100 ml. A nonsignificant increase in transtent with the pharmacologic action of as- fusions at delivery was also seen for those pirin in irreversibly damaging platelets, 81 women who used salicylates one to six which have a 10-day life span (42). Women times per week; the overall transfusion rate who took aspirin 6-10 days before delivery of 9 percent among salicylate users is comwould have had time to at least partially parable to the 10 percent observed by replenish their supply of platelets, enough Stuart et al. (34). The doses ranged from 2 to avoid excessive bleeding at delivery. to 12 tablets or powders per day of use. Women who used aspirin within 5 days of Although the results on habitual users or immediately after delivery would be are in agreement with the findings from more likely to experience abnormal bleed- other studies, the subjects, interviewers, ing, as most of the circulating platelets are and clinicians were aware of the purpose of still affected. the investigation, which was to examine The reasons for aspirin use were not effects of prolonged salicylate use on preggiven, leaving open the possibility of con- nancy, and, more importantly, the exposed

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

ASPIRIN USE IN PREGNANCY

women were selected based on their habitual use of salicylates in powders or tablets containing aspirin, phenacetin or salicylamide, and caffeine. Many had become habituated in their teens, and 45 percent of the daily users had taken them for at least 10 years. These women smoked more than controls, and were more likely to have a urinary tract infection during pregnancy. Neither factor was controlled, nor could this study separate the effects of aspirin from the other ingredients in headache powders. With respect to maternal bleeding, this report supports the findings of other, stronger studies. Four other studies using laboratory measures of hemostasis also support these findings. The first found that collagen-induced platelet aggregation was dysfunctional in maternal blood when aspirin had been taken at least once in the last week of pregnancy (33). Aggregation was present in all unexposed women and in women exposed to aspirin between 1 and 3 weeks prior to delivery. Aspirin exposure information was based on a daily diary kept by the mother during the last month of pregnancy (5) and not opened until after the delivery and laboratory tests had been conducted, thus assuring blind assessments of all outcomes. In a second study (43), 8 of 10 women who took aspirin, but no other drugs in the week preceding delivery and no systemic analgesics during labor, had severely impaired platelet aggregation (> 2 standard deviations below the mean of 80 percent aggregation among unexposed women). No doses were reported in this study, nor were any clinical parameters such as amount of blood loss. In the third study, aspirin was administered intravenously to prevent premature labor (44). Prothrombin (clotting) times were 30 percent longer than normal in blood from treated patients; however, no controls were used in this study. The fourth study was designed to examine effects of very low doses of aspirin on maternal and neonatal prostaglandin F l a thromboxane B 2 and platelet aggregation,

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

117

and on several measures of neonatal circulation (45). Doses were 20, 60, and 80 mg/ day beginning at approximately 37 weeks of gestation. There were 10 normotensive primigravida women in each dose group and in a placebo control group. Statistically significant reductions in maternal thromboxane B2 levels were observed among women treated for 1 week with 60 or 80 mg/day of aspirin. Maternal platelet aggregation in response to collagen stimulation was inhibited by 60 or 80 mg/ day of aspirin. Other studies of very low doses of aspirin (60 to 150 mg/day) for the prevention of pregnancy-induced hypertension did not examine laboratory measures of maternal hemostasis, but did report no abnormal bleeding during labor and delivery (13-16, 45). One of these observed longer bleeding times from incisions in the 28th week, which, although not statistically significant, represented a 32 percent increase in the aspirin treated patients as compared with placebo controls (15). Conclusion. The evidence from seven epidemiologic studies and from one study in rats is consistent in showing that aspirin consumption close to the time of delivery is associated with maternal hemostatic abnormalities. Clinically observable effects were seen at doses of three to five aspirin per day. Aspirin use earlier in pregnancy was not associated with maternal bleeding at the time of delivery, nor were doses up to 150 mg/day. Most of the studies did not report how they measured blood loss and it was likely to have been a subjective assessment. For this reason, the fact that investigators were blind to subjects' exposures in many of these studies (33, 34, 39, 46) strengthens their findings. Given the uncertainty of the timing of labor and delivery, it may be prudent to discourage aspirin use during the last trimester to reduce the likelihood of excessive bleeding. Exceptional circumstances, where prescribing low doses may be beneficial for specific conditions such as preeclampsia, are described below.

118

HERTZ-PICCIOTTO ET AL.

The clinical significance of an aspirin effect on maternal hemostasis has been questioned (47). Platelet aggregation and prothrombin times do not predict clinical bleeding very well. Clearly, though, an abnormal blood loss in 4 of 10 women who took aspirin within 5 days of delivery, compared with none among 34 unexposed women (34), is a substantial difference. Abnormal blood loss included frank postpartum hemorrhage, excess intraoperative bleeding during a cesarean section, and a large labial hematoma requiring evacuation. One of the women who took aspirin either in the immediate postpartum period or in the previous 5 days had such heavy blood loss as to require a transfusion (48). In general, however, the quantities of blood lost were not life threatening. Clinically observable effects on maternal hemostasis were seen at doses averaging 1,500 mg (less than five normal strength or three extrastrength aspirin tablets) per day (34), but other studies reporting such effects involved much higher doses. Platelet dysfunction was seen at much lower doses, e.g., 60 mg/day (46). In the data currently available, moderate increases in the risk of overt excess bleeding have not been observed at very low doses, though small increases in either the risk of hemorrhage or in the amount of blood lost could not be ruled out. Neonatal hemostatic abnormalities Animal data. Both aspirin and salicylate cross the placenta to reach fetal circulation and salicylate is likely to be present at higher plasma concentrations in the fetus than in the mother. Studies using mice have found increased hemorrhaging in the offspring of rodents receiving salicylates while pregnant. Following maternal injection of sodium salicylate either 1 or 2 days (but not 3 days) prior to anticipated delivery, neonatal subdural, gastrointestinal, and liver hemorrhage were observed in mouse pups (49-50). The occurrence of hemorrhage was accompanied by mortality of pups as high as 70 percent at maternal doses of 10 mg/20 g; however, there was

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

little opportunity to observe maternal toxicity because dams were sacrificed shortly after treatment. Similar results were obtained with orally administered aspirin or with sodium salicylate in mice (51). Subdural hemorrhage was also reported in ferrets treated by gavage with sodium salicylate (52) and in hamsters treated by gavage with salicylamide during embryogenesis (53). Experiments on mice examined the effects of sodium salicylate exposure on maternal and fetal clotting times. Compared with controls, prothrombin times in treated animals were not increased for dams or for fetuses showing no hemorrhage, but the mean prothrombin time of treated fetuses who showed subdural hemorrhage was about twice the mean control value (54). Animal studies demonstrating subdural hemorrhage, and also gastrointestinal and liver hemorrhage, in fetuses after exposure to salicylate suggest that metabolized aspirin (e.g., salicylate) may be the causal agent. This runs counter to the assumption that because aspirin has a short half-life, and because its metabolite salicylate is not an effective anticlotting agent in adults, the effect on neonatal hemostasis should be negligible. In addition, the observation of increased clotting times in the mouse fetus, but not in the mouse dam, suggests that clotting in newborns is in some way uniquely susceptible to salicylate. Determination of whether salicylate or aspirin is the active agent in neonatal bleeding disorders is important in determining clinical interventions. Human data. A case series of newborns exposed to aspirin late in gestation and having abnormal platelet aggregation has been published (55), as have case reports of purpura and life-threatening gastrointestinal hemorrhage (56-57). The main support for an aspirin effect on neonatal bleeding comes from several epidemiologic studies of infants whose mothers took aspirin within a short time before delivery (table 2). One study examined the effects of aspirin

ASPIRIN USE IN PREGNANCY

taken within 1 week of delivery on bleeding among premature (34 weeks gestation or less) or very low birth weight (< 1,500 g) infants (39). (Small infants are already at high risk of developing intracranial hemorrhage. For instance, 25-39 percent of such premature or very low birth weight infants experience peri- and intraventricular hemorrhage (58).) In the aspirin study, exposure information was obtained from the mother 2 days after delivery. Subsequently, the diagnosis of intracranial hemorrhage was made by computed tomography (CT) performed 3-7 days postdelivery. Infants exposed to aspirin in utero had a significantly higher incidence of intracranial hemorrhage (12 of 17, or 71 percent) than those exposed to neither aspirin nor acetaminophen (31 of 71, or 44, percent, one-tailed p < 0.05). Although the indications for drug use were not provided, the acetaminophen group differed very little from the control group. Given that both drugs are used largely for treating the same symptoms, it is likely that aspirin, rather than the condition for which it was taken, was responsible for the increase in intracranial hemorrhage. Those reading the CT scans were blind to the exposure status of the neonates (C. M. Rumack, University of Colorado School of Medicine, personal communication, August 1,1990). The doses in this study were low, a total of one to five tablets over the course of a week. The very low birth weight group represents about 1 percent of live births in the United States, or approximately 40,000 babies per year (59). Thus, the findings are of public health significance. Given that these infants are already at high risk of intracranial bleeding, can the aspirin findings generalize to normal, full-term newborns? Several studies confirm an excess of bleeding abnormalities in normal full-term infants exposed to aspirin prenatally (table 2). In a study of normal, full-term pregnancies, Stuart et al. (34) observed that 9 of the 10 infants whose mothers took aspirin 0-5 days before delivery had clinical signs of bleeding tendencies (including petechiae,

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

119

cephalhematoma, circumcision bleeding, etc.). Only 1 of the 34 nonusers of aspirin had signs of bleeding. The risk ratio of 30.6 (or 15.8 after adjusting for small sample size) was highly significant (p < 0.000001), although none of the observed manifestations of bleeding are life threatening. None of the seven infants whose mothers took aspirin 6-10 days prior to delivery had bleeding abnormalities. All of the clinical examinations were conducted by observers blind to the exposure status of the babies. Another study of full-term infants born to mothers with normal pregnancies and uneventful deliveries found differences in bleeding, in platelet aggregation, and in coagulation factor XII comparing aspirinexposed with unexposed infants (33). In this prospective study, mothers kept diaries of all drugs used during the last month of pregnancy. The diaries were not opened until after laboratory tests and clinical examination of the newborns were conducted. Of 14 newborns whose mothers used aspirin during the week prior to delivery, three cases of bleeding occurred (one gastrointestinal, one cephalhematoma, and one bilateral periorbital purpura), compared with one among the 17 newborns whose mothers did not take aspirin during the last 3 weeks of pregnancy. The risk ratio of 3.6 (1.9 after adjusting for the small sample) was not significantly different from 1.0. Laboratory tests indicated that collageninduced platelet aggregation was absent in all 14 exposed neonates, while it was normal in all the unexposed neonates. Coagulation factor XII activity was diminished in the exposed group compared with the unexposed group (p < 0.005). These findings (33) are notable for several reasons: 1) the effects are marked, especially for the laboratory indices, in spite of a small sample size; 2) the doses, ranging from one aspirin per week to four aspirin per day, were not unusually high as compared with typical home use; and 3) laboratory and clinical evaluations were conducted blindly with respect to the exposure status of the study subjects, even though 4)

120

HERTZ-PICCIOTTO ET AL.

exposure information was collected prospectively. Corby and Schulman (43) also confirmed a striking deficiency in collagen-induced platelet aggregation among full-term, normal birth weight neonates exposed only to aspirin in their last week of gestation. In addition, these authors reported an in vitro test in which aspirin was added to blood from an unexposed mother and her infant; the suppression of platelet function in blood from the newborn was 10- to 15-fold greater than in maternal blood. It appears that full-term neonates may be far more susceptible to the effect of aspirin on platelets than normal adults. No hemorrhagic complications were noted in five studies involving very low doses of aspirin ranging from 20 to 150 mg/ day (13-16, 46). In one of these studies (16), women had been instructed to discontinue aspirin 10 days before their due date, but in the others treatment continued to delivery. In the one very low dose study which examined neonatal platelet aggregation (46), aspirin-exposed neonates did not differ from placebo controls, but the number of treated neonates evaluated for this outcome was very small (four or five in each dose group, for a total of 14), and compliance rates were not reported. This study did note a statistically significant reduction in platelet thromboxane B 2 production among neonates whose mothers took 20 or 60 mg/day of aspirin. Conclusion. The epidemiologic data provide consistent evidence that the risk of bleeding in the newborn increases when maternal exposure to aspirin at analgesic/ antipyretic doses has occurred shortly before delivery. Platelet aggregation in neonates without drug exposure is already much lower than in adults (43, 46). The pharmacologic effect of aspirin in inhibiting prostaglandin synthesis, and thereby interfering with platelet aggregation, combined with the diminished capability of the newborn to metabolize and eliminate aspirin, provides a plausible biologic mechanism for increased neonatal bleeding.

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

Taken together, the findings in mice, ferrets, and hamsters, the epidemiologic studies, and the pharmacologic data support a causal inference. Effects on very small and premature infants were observed at doses totaling one to five aspirin (325-2,500 mg) in the week before delivery; clinical signs of bleeding in full-term infants appeared at doses averaging 1,500 mg/day, but not at doses < 150 mg/day. A small effect at these very low doses could not be ruled out. Larger studies of aspirin therapy for preventing preeclampsia will provide a more complete picture of the effects of these very low doses. Again, given that 8.9 percent of infants are born at least 3 weeks preterm (60) and are, therefore, particularly vulnerable to intracranial bleeding, aspirin should be taken in the last trimester of pregnancy only if prescribed for a specific reason. Prolonged gestation and labor

In midpregnancy, prostaglandins have a role in increasing smooth muscle tone, and at term, they influence the ripening and effacement of the cervix and stimulate rhythmic uterine contractility (61). Aspirin, through its inhibition of prostaglandin synthesis, may, therefore, influence the onset and length of labor. Because some of the animal studies do not clearly distinguish between an effect on length of gestation and an effect on length of labor, we combine these end points for the review of animal results, then discuss the epidemiologic data on the two end points separately. Animal data. Studies in rats have demonstrated an effect of aspirin in prolonging gestation and/or labor (40, 62-64). In each study, indomethacin, a more potent prostaglandin inhibitor than aspirin, produced a similar effect. Two studies recorded only total gestation length; thus, the effect could be because of delays in the initiation and/ or increases in the duration of labor (62, 64). In two studies of rats, aspirin increased labor from a normal duration of 2.5 hours

ASPIRIN USE IN PREGNANCY

to 3-4+ hours (40, 63). Doses were 60 mg/ kg/day on days 18-21 (40) or 200 mg/kg/ day from day 15 on (63). The investigators indicated that in some animals, labor was dysfunctional and animals sometimes had to be killed when it became apparent that parturition could not be accomplished. However, dose-response data indicate that doses which do not increase dam mortality will extend labor and/or gestation (64). While information is not available for species other than rats, prostaglandins are known to play a role in the initiation of labor in many species, and indomethacin has been shown to prevent labor in monkeys (65). Prolonged labor, human data. Three epidemiologic studies (7,37,46) examined prolonged labor in relation to aspirin use (table 2). Lewis and Schulman (37), in reviewing 20 years of obstetric records, found that women who took high therapeutic doses of aspirin for rheumatoid arthritis during pregnancy had about 70 percent longer labor (mean = 12.1 hours) than either of the two control groups (mean = 7.3 hours for unexposed rheumatoid arthritis patients, mean = 7.0 for healthy unexposed women), both differences highly significant (p < 0.005). Although the definition of onset of labor may not have been consistent, 18 of the 103 aspirin-treated women had labor lasting over 24 hours, compared with none lasting over 24 hours in the unexposed women. The difference in mean parity between exposed women (1.37) and the two unexposed groups (1.50 for diseased, 1.62 for nondiseased) was too small to account for the large difference in the duration of labor. This study did not control for potential confounders such as other drugs or the severity of disease. While such factors were unlikely to be related to the bleeding outcomes, they could have influenced prostaglandin-related effects such as length of labor. The study of habitual aspirin users did not find any significant differences between aspirin users and nonusers in the duration of labor (7). This analysis was restricted to

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

121

the approximately 60 percent of the study population who were multiparous and had uncomplicated deliveries. If prolongation of labor is associated with other complications, or if labor in primiparous women requires greater prostaglandin stimulation of the cervix than in multiparous women, then an effect of aspirin could be missed by such exclusions. The very low dose trial of primigravid women found no differences in the length of labor between placebo controls and any of three dose groups treated with 20, 60, or 80 mg/day of aspirin (46). Prolonged gestation, human data. Two studies found increases in the length of gestation among aspirin-exposed women. Women whose labor was induced or who had a cesarean section before term were excluded, either from the study (37) or for the analysis of this outcome (7). Lewis and Schulman (37) found that women treated with aspirin for rheumatoid arthritis had pregnancies that were, on average, 1 week longer than those of women with or without arthritis who took no aspirin (p < 0.025). No significant differences in length of gestation were found between the two control groups. With a cut point of 42 weeks, 42 percent of the aspirin-treated group versus 3 percent of the two control groups combined had postmature pregnancies, yielding a risk ratio of 14 (p < 0.001). The pregnancies of women habituated to salicylate-containing powders (both daily and one to six times per week) lasted, on average, 1 week longer than those of nonusers (39.7 vs. 38.7 weeks, p < 0.05) (7). The risk of postmaturity for each of the two exposure groups was 16 percent while the risk for the unexposed group was only 4 percent (p < 0.02 by our calculations). A related finding was published by Wolff et al. (44), who reported administering aspirin to successfully delay the onset of premature labor, but did not use a control group. The doses apparently caused numerous side effects. Among the three groups of 10 normotensive primigravida women each receiving

122

HERTZ-PICCIOTTO ET AL.

very low doses of aspirin starting in week 37, no prolongation of gestation was observed and no pregnancy lasted more than 41 weeks (46). Conclusion. Prostaglandins are involved in the initiation and duration of labor, and in both term and preterm labor their concentrations increase in the amniotic fluid (66). Three inhibitors of prostaglandin synthesis, aspirin (26), salicylate (67), and indomethacin (68), all cause cessation of spontaneous term or preterm labor in women. Tocolytic doses (doses capable of stopping labor) in humans are approximately 100 mg for indomethacin, 1,000 mg for aspirin, and 2,000 mg for sodium salicylate. Thus, a delay in the onset of labor and an increase in its duration are both biologically plausible effects of aspirin. The evidence from women treated for arthritis suggests that the activity of aspirin in inhibiting prostaglandin synthesis may result in prolonging labor, and studies in rats support this conclusion. Confirmatory studies would be desirable. The only data adequate for evaluating doses less than 3,000 mg of aspirin per day is the very low dose study by Sibai et al. (46), who found no effect on length of labor. No data are available for this end point at usual analgesic or antipyretic doses in the 300-2,500 mg/day range. While the epidemiologic data from one study strongly suggest that continuous high doses of aspirin (> 3,000 mg/day), by inhibiting prostaglandin synthesis, may delay the time of birth, confirmatory studies are lacking. The experimental data on this end point are weak. If causal, this effect is of clinical and public health concern, since postmature infants have higher rates of perinatal mortality (69). However, no data are available to evaluate the effect on length of gestation of moderate, intermittent, or occasional aspirin use at adult analgesic or antipyretic doses. In one small study, very low doses (< 81 mg/day) did not prolong gestation in full-term pregnancies of normotensive women.

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

Low birth weight and perinatal mortality Animal data. Several animal teratology studies noted reduced fetal weights at the time of examination (late pregnancy, just prior to parturition) (52, 70-72). Growth retardation is commonly associated with disrupted development in studies where malformations occur, although one study demonstrated reduced fetal weight when a subteratogenic dose of aspirin (125 mg/kg/ day) was administered throughout gestation (73). Growth retardation is not typically a primary effect of in utero exposure to salicylates in animals. Pup mortality has been observed in connection with increases in length of gestation and/or labor (62) and in association with aspirin-induced hemorrhage (49-50). Human data. Several studies examined aspirin consumption in relation to low birth weight and perinatal mortality. Infants born to habituated daily users of aspirin/ caffeine combinations had significantly lower birth weight than infants of nonusers (mean of 3,283 g vs. 3,502 g) (74). Infants of mothers who used these drugs one to six times per week also weighed less (mean = 3,372 g) than unexposed infants, though not significantly so. There were five stillbirths, all born to aspirin users, but this finding could have been the result of chance (p = 0.44). When past and present pregnancies were combined, salicylate users showed a significant excess in stillbirths or perinatal mortality, but not in neonatal deaths alone (74). These findings could have been because of other characteristics of habitual aspirin users. For instance, the number of years of aspirin use prior to pregnancy was associated with reduced birth weight, suggesting a cumulative adverse effect on maternal health. Renal compromise is one such adverse effect known to occur in long-term users of analgesic mixtures (23). Also, although users and nonusers were matched for several determinants of birth weight, aspirin users smoked more than the controls, and the crude adjustment of adding

ASPIRIN USE IN PREGNANCY

180 g to the weight of infants born to smokers does not take into account the timing or amount of smoking. Caffeine itself has been reported to have an association with low birth weight, and other potential confounders, such as alcohol, were also not controlled. In contrast, among infants born to women with arthritis or other musculoskeletal disorders, there was no difference in mean birth weight between those exposed to aspirin and those unexposed (37). Infants of women without a musculoskeletal disorder had a higher mean birth weight than either group, suggesting that the disease itself, but not aspirin, influences birth weight. Additionally, mean birth weight was not associated with "heavy" or "other" aspirin exposure (see table 2 for exposure groups) in the large, prospective Collaborative Perinatal Project data (75) after stratifying on ethnicity. The lack of association in the unadjusted analysis was not altered by multivariate linear regressions that included the covariables duration of pregnancy, smoking, birth order, and preexisting hypertension. Similarly, neither stillbirths nor neonatal deaths were elevated among "heavy" or "other" aspirin users after standardizing for both cigarette smoking and parity. Among women at high risk of pregnancyinduced hypertension or placental insufficiency, very low prophylactic doses of aspirin improved birth weight, expressed either as means, or as the proportion with severe growth retardation (13-16, 76, 77). Stillbirths were also reduced for pregnancies of high-risk women treated with aspirin (13). Conclusion. Compared with the study on habitual aspirin/caffeine users (74), the two studies showing no effect of aspirin use on birth weight (37, 75) are likely to be less subject to confounding or other biases. One used multivariate control for other risk factors (75) and the other used a control group with similar medical conditions as the

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

123

aspirin-exposed group. Although it is possible that an aspirin-induced tendency toward longer gestations could mask an effect on birth weight, no adequate studies implicate aspirin as a cause of low birth weight or perinatal mortality, and among high-risk women, the opposite effect was observed. Given the possibility of confounding resulting from either the select nature of habitual users or to conditions caused by their habit, any results on such a population that are not corroborated by other studies must be viewed cautiously. Preterm constriction of the ductus arteriosus During intrauterine life, the ductus arteriosus allows the fetal circulation to bypass the fetal lungs so that deoxygenated blood passes directly to the placenta, and oxygenated blood is shunted from the umbilical vein via the right ventricle and pulmonary artery to the aorta. At birth, the ductus arteriosus closes, forcing blood to pass from the right atrium to the lungs and from the lungs to the aorta. During fetal life, patency of the ductus arteriosus is maintained by prostaglandins, while cyclooxygenase inhibitors constrict it (61, 78). Constriction of the ductus arteriosus in fetal life has been postulated to lead to abnormalities in pulmonary vasculature that would promote pulmonary hypertension in the newborn (79-81). Animal data. In animals, a single high dose of aspirin late in pregnancy produces constriction of the ductus arteriosus through its inhibition of prostaglandin synthesis. This has been demonstrated anatomically in rats (82) and physiologically (by monitoring ductus arteriosus blood pressure) in sheep (79). The effect in rats was dose-dependent, with a slight constriction at "clinical" doses of 10 mg/kg and much greater effects at 100 and 1,000 mg/ kg (82). The effect was also produced by indomethacin, ibuprofen, and a number of other nonsteroidal antiinflammatory drugs. In sheep (79), 20 mg/kg of aspirin or acet-

124

HERTZ-PICCIOTTO ET AL.

aminophen resulted in an increased pressure gradient across the ductus arteriosus. This dose is equivalent to four analgesic tablets. Stenosis of the ductus arteriosus was also reported when aspirin was administered during embryogenesis (83). Chronic low doses of aspirin at any time during pregnancy may cause some constriction of the ductus arteriosus, but possible consequences of this partial closure have not been explored in animal models. Human data. There have been case reports in which salicylate ingestion was associated with preterm occlusion of the ductus arteriosus (80, 84). In one fetus, numerous abnormalities developed in the pulmonary vasculature, leading to neonatal death (80). In general, however, this condition would not be easy to diagnose, even with a routine autopsy. Thus, it is not surprising that epidemiologic studies have not examined maternal aspirin consumption in relation to premature closure of the ductus arteriosus. However, nonaspirin cyclooxygenase inhibitors have shown some association with persistent pulmonary hypertension in the newborn (17, 81), a condition which could be related to premature constriction of the ductus arteriosus (81). None of the neonates exposed in utero to very low doses (< 81 mg/day) starting at week 37 of gestation showed evidence of premature closure of the ductus arteriosus or elevation of pulmonary arterial pressure. However, the study groups were very small and these are rare outcomes, resulting in low power. Conclusion. Some clinical and animal data suggest a possible adverse effect on functional and morphologic development of the fetal vascular system. Whether or not aspirin has an etiologic role in either premature occlusion of the ductus arteriosus or in persistent pulmonary hypertension remains unclear. ADVERSE EFFECTS OF ASPIRIN CONSUMPTION AT AN EARLY OR UNSPECIFIED STAGE IN PREGNANCY

The most extensively studied effect of aspirin consumption early in pregnancy is

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

the risk of congenital malformations. Both animal and human studies report on teratogenic end points ranging from a general "all malformations" to very specific anatomically defined defects such as transposition of the great vessels. The other end point of concern has been cognitive and behavioral development. One study examined childhood cancers in relation to analgesics and antipyretics taken at any point in pregnancy. Congenital malformations Animal data. Salicylate teratogenicity in animals was first reported in 1959 (85). Administration of sodium salicylate by injection at doses up to 180 mg per rat on a single sensitive day during embryogenesis (day 9, 10, or 11) produced a malformation rate of 41 percent overall, 26 percent for external or skeletal defects. In the 484 controls, all organs were not examined, but the rate for external and skeletal malformations was zero. Methyl salicylate was also studied and led to 52 percent malformations at high doses. Some of these malformations rarely occurred spontaneously, including craniorachischisis and related neural tube defects (such as spina bifida, exencephaly, anencephaly, and hydrocephalus) and transposition of the great vessels of the heart (83). Aspirin has been used as a positive control in animal teratogenicity studies (86, 87). Further studies in mice (88) and rats (89) observed teratogenic effects when aspirin was administered orally. In animal studies, oral administration of aspirin is typically by oral intubation, thus simulating administration of a tablet or capsule. The study in rats also demonstrated transfer of salicylate to the embryo, and both doseresponse and stage-specific characteristics of aspirin-induced teratogenesis. The same study found that pretreatment of animals with benzoic acid significantly increased both the concentration of salicylate in embryos and the malformation rates (over the rates produced by aspirin alone). By competing for plasma protein binding sites, and by binding to metabolic enzymes, benzoic

ASPIRIN USE IN PREGNANCY

acid can produce elevated levels of salicylate and increase the duration of exposure for the embryo. A later study demonstrated that in animals treated with sodium salicylate, pretreatment with iron also produced significantly higher levels of embryonic salicylate, a significant increase in extra ribs over that induced by salicylate alone, and a nonsignificant potentiation of salicylate teratogenicity for other malformations (90). Subsequently, aspirin was used to probe the teratogenic sensitivity of several species including mouse (86), cat (71), dog (91), ferret (52), and monkey (92). Some evidence of teratogenesis and embryotoxicity was demonstrated in each of these species although effective doses and times of administration varied. Cardiac malformations were reported in the majority of animal teratogenicity studies where internal organs were examined (table 3), including six studies in rats (52, 70, 83, 87, 89, 93), and one each in ferrets (52), beagles (91), and monkeys (92), with one negative study in monkeys at a lower dose and later timing (94). In one study (89), 34 percent of rat fetuses examined had levorotation (rotation of the heart to the left), while the total malformation rate in untreated rats was 1 percent and in vehicle controls 4 percent. Four other studies in rats observed either displacement of the heart to the right (dextrocardia) or rotation to the left. Levorotation and dextrocardia both occur in humans and range from having severe sequelae to being asymptomatic and of little medical concern. Cardiovascular defects were reported in 30 of 159 (19 percent) rat fetuses examined after maternal treatment with sodium or methyl salicylate during embryogenesis (83), including dextrocardia, ventricular septal defect, transposition of the great vessels, persistent truncus arteriosus, and stenosis of the ductus arteriosus. Ventricular septal defect was induced by acetylsalicylic acid in beagles (91) and rhesus monkeys (92). Typically, doses of aspirin > 200 mg/kg/ day for 1-7 days were used in these studies, and maternal toxicity in the form of hyper-

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

125

ventilation, dyspnea, emesis, transient weight loss, and gastric ulcers was observed in some experiments. Several of these complications also occur in humans receiving high dose aspirin treatment. Less severe side effects that are common in humans, such as tinnitus and gastric irritation, would not be detected in animal studies. Several rat studies showing cardiac defects reported lower food intake and about a 1012 percent depression in weight gain among treated animals. Teratogenicity does occur at doses that do not induce maternal mortality or serious maternal toxicity. Human data on teratogenicity. Before presenting the epidemiologic studies, we note a reported case series of eight malformed infants whose mothers took aspirin daily in substantial doses at known time periods early in gestation (95). These are summarized in table 4. While case reports are always weak evidence because of the lack of a comparison group, these reports are notable for the relatively high quality exposure data, unequaled in any of the epidemiologic investigations of congenital malformations. It is interesting that anomalies of the hands and feet, which appear in this case series, were also detected in the Collaborative Perinatal Project data (1) and in animal experiments (96). Table 5 summarizes the epidemiologic studies examining the teratogenicity of aspirin. What is most striking is the poor quality of exposure data. No study collected the amount of aspirin taken. For most studies of malformations, exposure was categorized as "any" versus "none." Information on frequency of taking aspirin was collected for two study populations (1, 36, 97). In one of these, the "heavy" exposure category was essentially a low dose, i.e., 8 days/month in one of several months, which could be as low as eight aspirin in the whole month (1, 97). The most specific characterization of exposure was in Werler et al. (36) who categorized aspirin use as daily, 2-6 days per week, or < 2 days per week. Even these data are a far cry from the information in the case series. As for timing, three studies collected data by

TABLE 3

Cardiac defects produced by aspirin and other salicylates in animal teratology studies* Study (reference)

Species

Effective exposuret

Description of cardiac defects

Incidence}

Reported maternal toxicity

Takacs and Warkany, Rats 1968 (83)

MeS, NaS, ASA§ subcutaneous 30 of 159 Dextrocardia Maternal mortality and fetal resorptions injection doses not specified Transposition of great vessels high among dams receiving NaS and Persistent truncus arteriosus MeS Stenosis of ductus arteriosus No mention of maternal toxicity for Ventricular septal defect!! dams exposed to ASA

Kimmel et al., 1971 (89)

Rats

ASA, 625 mg/kg day 10 of gestation

Butcher et al., 1972 (70)

Rats

ASA, 250 mg/kg days 8-10 of gestation

2 of 79 Levorotation of the heart

Not mentioned

Beall and Klein, 1977 (93)

Rats

ASA, 250 mg/kg days 7-10 of gestation

1 of 42

Dextrocardia

Lack of weight gain and reduced food intake during treatment period (but not afterwards)

Tanigawa et al., 1979 (87)

Rats

ASA, 225 mg/kg days 7-17 of gestation

2 of 83

Dextrocardia

Reduced weight gain and decreased food intake at start of treatment, returning to control level by day 14; greater water intake throughout treatment (10 g/day)

Gulamhusein et al., 1980 (52)

Khera, 1976 (71)

Rats

NaS, 400 mg/kg day 11.5 of gestation

Ferrets

NaS, 400 mg/kg day 13 of gestation

Cats

No effective dose; doses were up to 50 mg/kg, days 10-20 of gestation

Robertson et al., 1979 Dogs ASA, 400 mg/kg days 23-30 of (91) (beagles) gestation

18 of 531| Levorotation of the heart

Hyperventilation for a few hours after dosing; transient weight loss (2-10 g) within 2-3 days after dosing

2 of 59 Transposition of great vessels No obvious weight loss or reduced food Heart defect (no details intake, though these were not specifigiven) cally monitored 1 of 4 Transposition of great vessels

0 of 52 None

No "apparent signs of ill health or effects on pregnancy"

6 of 16 Ventricular septal defect Hyperplasia of ventricular

Emesis mainly in those dosed on days 15-22 of gestation; less so in those

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

M

r

ASPIRIN USE IN PREGNANCY

§ I ••S -u

3§ 5

1

2 .5 o 3

1I*II o

o

c o > .He

o

g

13 9 >> a

gltiS

S &J3 g.,2 ft o> -2 S; ^3

(Ti

Q> 0) "S

•fi«s.i

S & * J? g a Q

»

o o

o

•a

a o c

I

CO JJ 2

~36

o c 2

s

§ a '-a

2

B c

CO

1 s 8 -g g R 3

II

1 i ^ g -° | a -S a t-

iijjip *

O5 C



|-5:I.3a|| .5

S N W C O H H

Tt

1/5 »H O

CN

H

INCO

CO

CO

Qj

CO

2

.g "* jg

Hi Q

])

C

„ a

v

I

8 "g 3 - g o

•2

3

I i:! § *° •a I1 11 o o

a>

8.*

*_

-1

* *° 2 -*J CJ ^

Ci*

I

8.

134

HERTZ-PICCIOTTO ET AL.

open-ended question was asked and accurate information on dose or time of use was not collected. Also, although confounding was ruled out from parity, maternal age, education, and insulin use, no multivariate analysis was performed to control for multiple drug usage. Comparisons showing no aspirin effect were not reported. Transposition of the great arteries was not significantly related to aspirin use in the unadjusted analysis of the Collaborative Perinatal Project data (1). The third study of congenital heart defects (35) investigated associations with the use of Bendectin (doxylamine succinate), aspirin, acetaminophen, and other drugs taken during pregnancy. The study group consisted of 298 mothers of infants with severe congenital heart malformations, and 738 mothers of a random selection of infants (see table 5). In addition, within the case group, those with specific malformations were compared with those having other cardiac malformations. Exposure data were collected in a structured telephone interview with the mother at a mean of 13 months after delivery, and supplemented by review of obstetric records. No doses were recorded. Exposure was defined as use versus nonuse during the first trimester (14-92 days after the last menstrual period). Aspirin was significantly associated with several malformations. Combining transposition of the great arteries, tetralogy of Fallot, aortic stenosis, truncus arteriosus, double-outlet right ventricle, and hypoplastic left ventricle, the estimated relative risk was 2.1 (90 percent confidence interval 1.1-3.9), where all other cardiac cases served as the comparison group. After controlling for other drugs and symptoms (including febrile illnesses) using logistic regression, four specific malformations were found to have an elevated risk: aortic stenosis, hypoplastic left ventricle, transposition of the great arteries, and coarctation of the aorta, of which hypoplastic left ventricle and coarctation of the aorta remained statistically significant (table 6). Tetralogy of Fallot was not reported separately.

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

The fourth and most recent data bearing on aspirin's relation to these defects (36) come from a case-control study using cardiac defects as cases and noncardiac birth defects as controls, all identified from 1976 to 1986 through a birth defects surveillance program (see table 5). Exposure data were collected through a structured home interview approximately 4 months after delivery. Exposure was defined as aspirin use in the first 3 lunar months, a period which encompasses the time when cardiac morphogenesis occurs. Detailed frequency and timing data were collected as well as specific conditions for which each drug was taken. In multivariate analyses controlling for demographics, family history of malformation, diabetes, antibiotics, and other analgesics, no association was observed between aspirin use and all cardiac defects, or any of the following: aortic stenosis, coarctation of the aorta, hypoplastic left ventricle, transposition of the great arteries, or conotruncal defect. When analyses were done by frequency of use, there was no pattern supporting a dose response. Analyses controlling for timing of aspirin use and the conditions for which it was used were not reported. The use of antibiotics and other nonanalgesics was used as a surrogate for the conditions aspirin is used to treat. For all of the risk ratios shown in table 6, reporting bias was controlled either through prospective collection of exposure data (1) or through the use of other "diseased" subjects as controls (8, 35, 36). While the most recent study appears to be inconsistent with the others, for only one of the defects examined in more than one study do the 90 percent confidence intervals fail to overlap (transposition of the great arteries). For the other defects examined in more than one study, confidence bands do overlap, i.e., all studies are consistent with an elevation in the risk ratio of about 1.3 (hypoplastic left ventricle), 1.2 to 1.4 (coarctation of the aorta), and 1.0 to 2.1 (aortic stenosis). The point estimates calculated in three studies suggest the possibility of an elevated risk for five cardiac

ASPIRIN USE IN PREGNANCY

defects, yet one study shows no evidence of any elevated risk. That this negative study had the largest number of cases and, among the case-control studies, the shortest time interval between pregnancy and interview argues strongly for its validity. Additionally, the use of home interviews might yield higher quality data than telephone interviews. Nevertheless, in all of the studies, downward biases would be expected because of inaccuracies of aspirin reporting and the use of a wider time interval than the true period of organogenesis of the specific cardiovascular structures. The lack of a dose-response (36) is not a convincing argument against a causal hypothesis since detection of a trend might well be hampered by inaccuracies in the exposure measure and by the small number of exposed cases in the high-dose category. Conclusion. The evidence for aspirin teratogenicity in animals is definitive. However, both the biochemical mechanism and the implications for human exposure are less clear. With regard to the relevance to humans, the teratogenic dose levels in the animal studies caused maternal toxicity in some studies, but it should be noted that humans take therapeutic doses of aspirin which produce toxic effects such as gastrointestinal bleeding. Second, in some studies, nothing more than a transient loss of appetite and a small, reversible weight loss occurred in dosed animals. Third, equivalent effect doses between species have not been established. Fourth, the organ site of action is not necessarily predictive from animals to humans. Nevertheless, there are at least three species in which aspirin has been associated with cardiac defects closely related to those associated with aspirin in several human studies. With respect to the mechanism, salicylic acid (rather than aspirin itself), or a metabolic product of salicylic acid, has been shown to be the teratogenic agent in animals (89, 102). In pharmacokinetic studies, maternal and fetal plasma levels of salicylate are good predictors of adverse fetal outcomes in rats (25). However, since salicylic acid is only a very weak inhibitor of

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

135

prostaglandin synthesis (22), this pathway appears not to be the mechanism of action. Indeed, a series of studies by Klein et al. (27) did not support the hypothesis that prostaglandin synthesis is the teratogenic mechanism. Teratogenic action of other prostaglandin synthesis inhibitors would also bear on this issue. Reports on teratogenic action of indomethacin are not consistent; poor maternal-fetal distribution of indomethacin during early pregnancy (96) may contribute to these discrepancies. How likely is it that aspirin is a human teratogen? At this point, the data are inconclusive. Despite strong evidence from an extensive animal literature, most epidemiologic investigations have been hampered by inadequate exposure information, many by insufficient control for potential confounders, and some by inadequate safeguards against reporting bias. The cardiac defects are the only ones that have been pursued, and here the findings are also inconclusive. Exposure to aspirin early in pregnancy is not uncommon (table 1). Daily use of aspirin in the first trimester was reported by 3 percent of mothers giving birth between 1976 and 1986 to infants with either a cardiac or other defect. In the southern United States, "headache" powders with salicylate/caffeine combinations are used as they were in Australia. For these women or other regular users of aspirin, the teratogenic risk is not clearly known. The data are insufficient to rule out a substantial risk associated with moderate (analgesic/ antipyretic) doses during a critical time period in gestation. Additional possible reasons for concern are the apparent potentiation of aspirin teratogenicity in animals by both iron (90), which is taken by most pregnant women (4, 5), and benzoic acid (89), a common food additive (89, 103). Cognitive development Behavioral teratology or psychoteratology attempt to relate functional impairment in postnatal development or in adult life with insults to the fetus or embryo. The central nervous system may be sensitive to

136

HERTZ-PICCIOTTO ET AL.

exposures at dose levels too low to cause structural anomalies, yet high enough to be manifest as deficits in cognitive or behavioral development. Prenatal lead exposure, for instance, has been associated with deficits in mental development through at least 2 years of age (104, 105), but it does not appear to be associated with major malformations (106). Animal data. Changes in behavior of immature and adult rats after exposure to aspirin via treatment of dams during embryogenesis have been reported (70, 72, 90). Doses too low to produce gross central nervous system malformations were chosen in order to detect possible effects on central nervous system function. In one study (70), adult rats prenatally exposed to 250 mg/kg of aspirin on days 8, 9, and 10 showed increased errors in learning and reversal of a water maze task at a dose which produced no external malformations. This was a carefully controlled study; sex differences were analyzed and cross-fostering procedures were used to ensure that in utero exposure to aspirin, not differences in the behavior of the dam, was responsible for the learning deficits. In a second study (90), a lower dose (one injection of 250 or 500 mg/kg) of aspirin administered after pretreatment with ferrous gluconate produced a similar maze learning deficit in adult offspring. The hypothesis that the chelate formed from the iron salt would slow down metabolism of salicylate and facilitate its transfer to the fetus by increasing solubility was supported by data on salicylate concentrations in embryos. Other adverse effects observed in rats were increased activity in an open field among adult offspring exposed prenatally (87, 90) and abnormalities in the performance of simple behavior patterns, such as pivoting, olfactory orientation, and swimming among young preweaning rats (72). Taken together, these experiments suggest that aspirin may affect neurobehavioral development when administered to rats during embryogenesis. An interpretation of a direct effect on the central nervous

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

system is complicated by occurrence of growth retardation in many studies. Notably, studies of behavior in relation to postembryonic aspirin exposure have not been conducted and studies in species other than rats are lacking. Human data. In humans, two reports have addressed the question of whether prenatal exposure to aspirin in the early stages of gestation affects cognitive function in childhood. In the first (9), 1,529 pregnant women receiving prenatal care in 1974-1975 were interviewed in the fifth month of pregnancy, and 452 of the children had their intelligence quotient (IQ) determined at 4 years of age using the Weschler Preschool and Primary Scale of Intelligence Test. The primary hypothesis of this study was that prenatal exposures to alcohol would affect child development. Associations with aspirin and with acetaminophen use in the first half of pregnancy were examined. Questions included the onset, duration, and frequency of each drug consumed, as well as the reasons for their use. Exposure categories were none, one per month, two per month, one per week, several per week, daily, and several per day. Aspirin consumption was reported by 46 percent of the women and acetaminophen by 44 percent; these figures include 13 percent who used both. The reasons for their use were similar, mainly for headache, and, less frequently, for pain or infection. The statistical analysis controlled for parental education, mother-child interaction, birth order, alcohol, nutrition, preschool experience, antibiotic use, and other potential confounders. In utero aspirin exposure at least several times per week was associated with a 10 point lower mean IQ for girls at age 4 years, and a 1.3 point lower mean IQ for boys of the same age. Aspirin exposure was also related to children's attention decrements. This was measured by a test of the child's vigilance in responding to a screen on which visual images appear at random times. Assuming that acetaminophen use is not correlated with determinants of IQ, the credibility of these findings

ASPIRIN USE IN PREGNANCY

137

may be strengthened by the lack of associ- different tests for measuring IQ. Given the ation between acetaminophen and either careful control for confounding, the proIQ or the attention test. No decrements in spective collection of exposure information, physical growth parameters were found for and animal data showing an effect of aspieither group. rin on behavioral development, the findings After these results were published, an- of Streissguth et al. (9) cannot be dismissed other group of investigators tested the as- easily. Nevertheless, the interaction of aspirin/IQ hypothesis using nearly 20,000 pirin and sex of the child is not easily children from the Collaborative Perinatal explained, and could be an argument for Project (107); the mothers of 10,159 chil- chance fluctuations. No physiologic mechdren reported consumption of aspirin at anism has been demonstrated in animals, least once in the first 20 weeks of gestation but there is biologic plausibility based on whereas the mothers of 9,067 children did alterations in vascular flow induced by innot report any aspirin consumption during hibition of prostaglandin synthesis (81). gestation. Exposure was coded as frequency More specifically, fetal circulation to the of aspirin use: none, 1 day/month, 2-7 brain and certain other organs is decreased days/month, and 8 or more days/month. by prostaglandin inhibition (108). Confir"Dose" groups were formed by summing mation in other studies is needed. across the 5 months to create a crude index Childhood cancer of exposure. IQ was measured by the Stanford Binet Test. In a recent study investigating the relaThe statistical analysis controlled for tion between childhood cancers and the use some potential confounders, including ma- of drugs during pregnancy (109), prenatal ternal education and IQ, birth weight, and use of antipyretics and analgesics was socioeconomic index. In one of several found to be associated with subsequent regression models, children with major neoplasms in the offspring (RR = 1.36), in malformations, low birth weight, or syn- particular solid tumors (RR = 1.51). Childromes featuring developmental delays dren who died of cancer (n = 8,059) were were excluded. No significant association identified through the Oxford Survey of was found between aspirin use and IQ at Childhood Cancers. Prenatal information age 4 years, and, if anything, there was a regarding maternal illnesses and drug use slight, but not significant, trend toward a were obtained via interviews with mothers protective effect of aspirin. Attention span and review of medical records. was not evaluated. The statistical analyses controlled for illConclusion. While neither of the epide- nesses, exposure to x-rays, drugs, and somiologic studies was designed to test the ciodemographics, but no information was hypothesis of aspirin and IQ, both studies collected on maternal smoking, drinking, collected exposure data prospectively and or other environmental exposures, nor on both controlled for some of the determi- drug exposures or illnesses of the children. nants of IQ. The positive study by Recall was sometimes as long as 10 years Streissguth et al. (9) had several advantages and the low rate of aspirin reporting among over the Collaborative Perinatal Project controls (< 0.5 percent) suggests severe study: 1) more determinants of IQ were underascertainment of this exposure. Asmeasured (R2 = 0.62 vs. 0.32 in the Collab- pirin was not analyzed as an individual orative Perinatal Project), 2) more detailed medication but as part of the broad cateexposure data were collected, and 3) the gory of analgesics and antipyretics. When authors controlled for the conditions the drugs were categorized by type of metprompting women to use aspirin by having abolic reaction (e.g., oxidation, reduction, a comparison group of women exposed to acylation, etc.) none of the pathways acetaminophen. The two studies also used thought to be related to carcinogenesis were

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

138

HERTZ-PICCIOTTO ET AL.

significantly associated with either reticuloendothelial system neoplasms or with solid cancers. In spite of elaborate attempts to separate illnesses from the drugs taken for them, weaknesses in this study were too great to consider it supportive of a transplacental carcinogenic effect of aspirin. BENEFICIAL EFFECTS OF ASPIRIN SPECIFIC TO PREGNANCY

Low-dose aspirin has been shown to have beneficial effects on women who are at risk for pregnancy-induced hypertension and preeclampsia (hypertension plus proteinuria or edema) and on their offspring. It has also improved the outcomes for fetuses subjected to umbilical placental insufficiency or to presence of maternal lupus anticoagulant. With respect to these beneficial effects, dose and timing are critical. Pregnancy-induced hypertension The mechanism leading to pregnancyinduced hypertension is thought to involve an imbalance in the ratio of two vasoactive prostaglandins which exert opposing effects on the small arteries and on platelets. These are thromboxane A2, a vasoconstrictor and a promoter of platelet aggregation, and prostacyclin, a vasodilator and an inhibitor of platelet aggregation. Both are derived from arachidonic acid through enzymatic action of cyclooxygenase. In normal pregnancy, production of both is increased, and sensitivity to angiotensin II decreases. However, in women with pregnancy-induced hypertension, the increased biosynthesis of thromboxane A2 appears not to be accompanied by a parallel increase in prostacyclin (110-112). The rise in thromboxane may also be exaggerated for preeclamptic women (110). Under these conditions, some women experience increased sensitivity to angiotensin II (113) and blood flow to the uteroplacental unit may be reduced. Researchers are not in agreement as to whether the site of thromboxane biosynthesis during pregnancy is in the platelets or the placenta (114, 115).

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

Observational data (116) as well as clinical trials (13-16) indicate improved obstetric outcomes for women at high risk for pregnancy-induced hypertension who took low doses of aspirin. Crandon and Isherwood (116) found that among 98 primigravida women with no drug use during pregnancy, 16 percent had preeclampsia compared with 4 percent of 48 women who had taken aspirin or aspirin-containing compounds more than once every 2 weeks during pregnancy. Four randomized clinical trials using women at high risk for pregnancy-induced hypertension are summarized in table 7. Beaufils et al. (13) treated women with a regimen of dipyridamole and aspirin, or with placebo, and found a marked difference between the two groups. Also, major complications (see table 7) occurred in 18 percent of the placebo group but in none of the treated women. Benigni et al. (15) randomly allocated highrisk women to aspirin or placebo and also looked at a third group of normal pregnant women. Besides a marked reduction in pregnancy-induced hypertension, this study also found a significant improvement in duration of pregnancy (a mean of 39 weeks in treated women vs. 34 weeks in controls, with low-risk pregnant women having a mean of 40 weeks), and birth weight (2,922 g vs. 2,264 g, and 3,492 g in low-risk pregnancies). The study by Beaufils et al. (13) was not double blind, and Benigni et al. (15) do not specify this aspect of the protocol, but two trials that did report a double-blind protocol also showed dramatic improvements for women treated with aspirin. Wallenburg et al. (14) observed mild pregnancy-induced hypertension in 10 percent of the treated women, while hypertension-related problems ranging from mild pregnancy-induced hypertension to outright eclampsia occurred in 52 percent of those receiving placebos. Schiff et al. (16) also used a doubleblind design, with similar results. Hypertension with either severe proteinuria or low platelet count occurred in 3 percent of the aspirin group and 23 percent of the

TABLE 7

Clinical trials of low dose aspirin to prevent pregnancy-induced hypertension High risk women

Aspirin Study (reference)

Double blind

Dose

Timing

Treatment Total no.

Beaufils et al., 1985 (13)

No

Benigni et al., 1989 (15)

? Not stated

60 mg/day

12th week to delivery

17

Wallenburg et al., 1986 (14)

Yes

60 mg/day

28th week to delivery

21

Schiff et al., 1989 (16)

Yes

28-29th week to 10 days before due date

34

150 rag/day 3 months to delivery

100 mg/day

%PIH*

10

12

* PIH, pregnancy induced hypertension.

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

Total no.

%PIH

Other outcomes improved °ng aspirin treated group

am

Total no. %PIH 50

52

Low risk women

Placebo

12

16

19

23

52

31

36

Stillbirth Neonatal death Severe growth retardation 11

Duration of pregnancy Birth weight Premature delivery Severe growth retardation Cesarean birth

104

Length of gestation Birth weight Length of hospital stay

C 1

2 z c

t

z

> z o

140

HERTZ-PICCIOTTO ET AL.

placebo group, and improvements were observed in other outcomes as well (see table 7). While each of the trials used a different criterion for determining who was at risk for pregnancy-induced hypertension, taken as a whole they provide consistent and fairly strong evidence for a beneficial effect of aspirin in prophylactically reducing pregnancy-induced hypertension. These studies also suggest that the usual fetal and neonatal sequelae of pregnancy-induced hypertension, including increased risks for stillbirth, growth retardation, premature delivery, and cesarean birth, may be reduced among women taking low doses of aspirin. The prophylactic doses used were 60,100, and 150 mg/day; these are low doses. Sixty milligrams is less than one "children's" aspirin tablet. Schiff et al. (16) suggest that the pharmacologically effective dose may be even smaller, perhaps 40 mg/day. Considering that a typical analgesic dose may be between 650 mg and 2,000 mg or more over the course of a day, the difference between the very low prophylactic dose used in these trials, and the usual selfadministered dose is between 4-fold and 30fold or more. Beaufils et al. (13) and Benigni et al. (15) administered aspirin as early as the start of the second trimester. The trials which began treatment only in the third trimester also showed strong reductions in pregnancy-induced hypertension. Schiff et al. (16) found that some women were already experiencing pregnancy-induced hypertension at 28 weeks, but since these women were excluded from the trial, it is not known whether they would or would not have benefited from aspirin treatment. Although the two studies showing a complete elimination of pregnancy-induced hypertension began treatment at the start of the second trimester, these two were not double-blind studies. More importantly, comparisons between the trials are not meaningful because different methods were used to identify the high-risk women.

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

Beaufils et al. (13) and Benigni et al. (15) used previous obstetric histories, Wallenburg et al. (14) conducted a test for response to intravenously infused angiotensin II, and Schiff et al. (16) screened by obstetric history and then conducted a "rollover" test. Of relevance to the issue of risk, Schiff et al. (16) recommended that women cease aspirin treatment 10 days in advance of their due date. The strong evidence of adverse effects from aspirin consumption late in pregnancy points to the need for a serious assessment of risks versus benefits. All four clinical trials (13-16) reported no excessive bleeding by the mothers and no hemorrhagic complications in the infants. Benigni et al. (15) reported 32 percent longer bleeding times from incisions at week 28. A study by Sibai et al. (46) (discussed in the section "Adverse effects of aspirin consumption late in pregnancy") administered very low doses to normotensive primigravida women starting in the 37th week of gestation to assess maternal and neonatal prostaglandins, platelet aggregation, several indices of neonatal circulation, and the lengths of gestation and labor. Most of these end points did not differ between those treated with aspirin and placebo controls. The statistically significant positive findings were an inhibition of maternal platelet aggregation and reductions in thromboxane B 2 production in maternal and neonatal platelets. Among circumcised babies, no complications occurred. With respect to the secondary measurements of hemostasis, the study sample was very small; laboratory measurements were available for only four or five neonates in each dose group. Taken with the clinical trials for prevention of pregnancy-induced hypertension and its sequelae, the study by Sibai et al. (46) is consistent with the conclusion that the risk for grossly observable hemorrhagic complications is not highly elevated at these very low doses. It will take larger trials to determine whether slight or moderate risks of bleeding are associated with very low doses of aspirin.

ASPIRIN USE IN PREGNANCY

Umbilical placental insufficiency Low-dose aspirin has also been used to treat placental insufficiency (76, 77, 117). Like pregnancy-induced hypertension, umbilical placental insufficiency is also believed to be caused by a thromboxane/prostacyclin imbalance. Wallenburg and Rotmans (117) administered aspirin to women with two or more previous pregnancies characterized by idiopathic fetal growth retardation and placental infarction. A drastic reduction was observed in fetal growth retardation (defined as below the 10th percentile weight for gestational age), both overall (13 vs. 62 percent, RR = 0.21) and severe, i.e., below the 2.3rd percentile (0 vs. 27 percent), comparing aspirin-treated to untreated pregnancies. Trudinger et al. (76, 77) diagnosed placental insufficiency in women who did not exhibit hypertension themselves using Doppler studies of women initially identified as being at high risk. Those with high systolic/diastolic ratios in the umbilical artery were selected for the randomized double-blind, placebo-controlled trial, and 150 mg/day of aspirin was taken by the treatment group. Birth weight, head circumference, and placental weight (an indicator of placental function) were all significantly improved in the aspirin-treated women who had high systolic/diastolic ratios (above 95th but below 99.95th percentile). No improvement was observed for aspirin-treated women in the extreme category of placental insufficiency, i.e., those for whom diastolic flow was essentially absent. The authors suggest that irreversible vascular damage to the placenta may already have occurred. Lupus anticoagulant with repeat abortions A third condition for which low-dose aspirin treatment has been proposed occurs in women with an autoantibody called lupus anticoagulant. Women with lupus anticoagulant frequently have a 92-100 percent risk of pregnancy loss, often with high risks in the second and third trimester, e.g., live births occurring in 13 out of 173 preg-

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

141

nancies of 49 women (118), in none out of 14 pregnancies of six women (119), and in one out of 31 pregnancies of eight women (120). (This antiphospholipid antibody is a misnomer on two counts: 1) it is more common in patients without systemic lupus erythematosus (118, 121) and 2) it is more frequently associated with thrombotic than hemorrhagic episodes (119).) In the group of six women with lupus anticoagulant, treatment with prednisone and 75 mg of aspirin daily enabled five to deliver a live child (119), and a similar treatment for the group of eight women with a consistent history of prior losses also led to five live births (120). Elder et al. (122) treated 42 patients for whom eight of 84 previous pregnancies had resulted in a live birth. Of the 42 women, 16 had systemic lupus erythematosus; under treatment with 75 mg/day of aspirin they delivered live births in 16 out of 19 pregnancies. In women at very high risk of pregnancy loss related to lupus anticoagulant, lowdose aspirin with or without prednisone may dramatically improve pregnancy outcome. Beneficial effects, conclusion Pregnancy-induced hypertension and preeclampsia are among the main causes of maternal morbidity during pregnancy, and pose a substantial risk of stillbirth, premature labor/delivery, and growth retardation. Four trials of aspirin have reported consistent, clinically and statistically significant improvement in pregnancy outcome for women at risk of pregnancyinduced hypertension. Women with uteroplacental insufficiency or lupus anticoagulant also clearly benefit from aspirin therapy, as do their offspring. In addition, high doses of aspirin have been used to successfully halt premature labor (31, 44). Nevertheless, both for future clinical trials and in clinical practice, the potential risks for increased bleeding, both maternal and neonatal, should be closely monitored. Careful screening of women to identify those at high risk for pregnancy-induced

142

HERTZ-PICCIOTTO ET AL.

hypertension or placental insufficiency is vention of pregnancy-induced hypertension needed. For the majority of women needing may give a clearer picture of whether small analgesic or antipyretic agents, there would increases in these abnormalities can be obappear to be no clear benefit from aspirin served at exposures of less than 100 mg/ during pregnancy that could not be ob- day. tained through other medications which From the viewpoint of mechanisms, the appear, at this time, to be safer (e.g., acet- data on mice have shown a response of fetal aminophen, which does not induce clot- prothrombin time in the absence of materting). Nevertheless, while the pregnancy- nal response (54). This observation is related risks of aspirin use cannot be ig- suggestive of a stronger effect in newborns nored, the risks associated with alterna- of salicylate (as opposed to aspirin) on tives to aspirin are largely unknown. platelet disaggregation than has previously been assumed. Another possible explanaDIRECTIONS FOR FUTURE RESEARCH tion is that in the fetus, aspirin itself is far There seems to be little doubt that inges- more potent than in adults and/or is metion of aspirin late in pregnancy increases tabolized to salicylate far more slowly. In bleeding tendencies in the mother and neo- humans, the effect of aspirin on platelet nate. Clinical bleeding appears to increase aggregation in neonatal blood was 10- to at doses of three to five tablets per day 15-fold stronger than in maternal blood (1,000-2,500 mg/day), and abnormalities in (43). Further experimental work is needed other measures of hemostasis, such as to determine the active agent in the fetus platelet aggregation, occur at less than 81 as well as the clinical significance of this mg/day. At least three areas need further difference. consideration. First, quantifying this relaSimilar questions can be raised with retion in terms of dose and timing has yet to gard to the prolongation of labor and gesbe accomplished. Thus, a major step would tation, even though the role of prostaglanbe to characterize the dose-response rela- dins (and by implication, aspirin) in these tion between aspirin intake and maternal processes is not as fully understood as is blood loss, hematocrit, risks of neonatal the interrelation among aspirin, platelets, hemorrhage, bleeding tendencies, and loss and bleeding. If aspirin does prolong gesof coagulation factors. A second critical tation, this effect is a clear benefit for preneed is the development of criteria for eval- eclamptic women who are at high risk of uating the clinical significance of each of delivering prematurely. Of concern is the these outcomes. Such criteria must take prolongation of pregnancy for the majority into account the physiologic profile of neo- of women who carry to full-term; the risk nates, and in postpartum women both factors for postmaturity remain to be clarphysiologic factors, such as rapidly chang- ified. More basic research on prostaglandin ing hormone levels, and psychologic fac- levels in postmature pregnancies may shed tors, such as the risk of postpartum depres- light on the suggested role of aspirin and sion. A third task is to identify subgroups other inhibitors of prostaglandin synthesis with bleeding tendencies who may suffer a in delaying labor. Baseline information on higher risk from aspirin-induced compro- length of labor and an understanding of the mise in platelet aggregation. factors besides parity that influence this parameter are needed. Research on some of these issues may prove difficult as aspirin use has declined Human data demonstrating constriction (12). Aspirin use in pregnancy may decline of the ductus arteriosus in relation to aseven further if the warning labels proposed pirin exposure may be exceedingly difficult by the US Food and Drug Administration to collect, certainly on an epidemiologic (123) become policy. On the other hand, scale. The suggestion that persistent pullarger trials of low-dose aspirin for the pre- monary hypertension in the neonate is a

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

ASPIRIN USE IN PREGNANCY

response to prostaglandin inhibition (84) is based only on case reports. Pulmonary hypertension in the neonate can, however, be more readily diagnosed than premature occlusion of the ductus arteriosus, and may, therefore, be amenable to epidemiologic investigation. More generally, the role of prostaglandin synthesis inhibition in fetal oxygenation has yet to be investigated either in humans or in animals. Existing work on teratogenicity suggests several unanswered questions. First, very little work has been done to describe the mechanism of aspirin-induced teratogenicity in animals. Whether inhibition of prostaglandin synthesis plays a role is unclear. Other proposed mechanisms are inhibition of mucopolysaccharide synthesis (124), enhanced production of a lipoxygenase pathway intermediate which is a vasoconstricting agent (22), and chelation of essential trace metals (125). The potentiation of aspirin teratogenicity by iron through chelation also deserves further investigation. Prenatal iron supplementation might be enhancing some of the adverse effects of aspirin. Such potentiation might apply to end points other than teratogenicity. Specific malformations deserving further study are the cardiac abnormalities listed in table 5, as well as anomalies of the alimentary tract, central nervous system, and hands and feet. Future research must avoid pitfalls with regard to the choice of appropriate comparison groups, collection of adequate data on confounders, particularly febrile illnesses, and proper analytic control for these variables. Most especially, careful attention must be paid to ascertainment of exposure, including documentation of dose and timing. Open-ended questions should be abandoned in favor of structured questions with product names (38). Similar recommendations would apply to any future research on transplacental carcinogenesis, although the existing data are far weaker than the evidence for teratogenicity. At present, the evidence of an effect on cognitive and behavioral development is inconclusive. Further epidemiologic re-

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

143

search might examine more thoroughly some of the more subtle developmental end points associated with prenatal and childhood lead exposure, such as ability to hear and perceive language, and ability to both mainta;n and shift attention (126). On the experimental level, consideration should be given to whether the reduction in blood flow to the brain would be sufficient to cause the observed effects, e.g., whether it is comparable to the hypoxia produced by maternal smoking, which some studies have found to be a behavioral teratogen (127). The use of low-dose aspirin prophylaxis holds great promise for preventing pregnancy-induced hypertension and preeclampsia, uteroplacental insufficiency, and other thrombotic disorders, and for thereby reducing the risks of stillbirth, premature labor/delivery, and growth retardation. Nevertheless, other nonpharmacologic prophylactic measures have also been suggested in the literature. Olsen et al. (128) reported observational data suggesting that a diet rich in fish oils may improve birth weight, and that part, but not all, of this improvement may be the result of prolonging gestation. Andersen and Andersen (129) noted that the birth weight difference reported by Olsen et al. (128) was seen at all gestational ages, and presented further data suggesting that such a diet may also lead to a lower risk of preeclampsia. Olsen et al. (128) cite both animal and human data showing that (n-3)-polyunsaturated fatty acids modify production of prostacyclins, and delineate three possible biochemical mechanisms for such an effect. These researchers suggest that marine oils may, like aspirin, alter the thromboxane/prostacyclin ratio in favor of the prostacyclin. A clinical trial of this simple nonpharmacologic intervention would seem to be in order. Because the timing of prophylactic administration of aspirin is in the latter part of pregnancy, the potential risks for increased bleeding, particularly intracranial hemorrhage, should be evaluated carefully

144

HERTZ-PICCIOTTO ET AL.

in any large trials, and the lowest pharmacologically effective dose should be used. Advising women to cease treatment a few weeks before their due date will protect against the bleeding effects, but only partially, since a substantial proportion of women deliver before term. Also, evidence suggests that there are at least two subgroups of women with pregnancyinduced hypertension, only one of which is responsive to aspirin (130-132). Determination of how these two subgroups differ, and the development of means to identify who is likely to respond, should be a high priority for further research. Even more fundamental is to determine the best method of screening for those who are at risk, given that most pregnant women are not at high risk of pregnancy-induced hypertension. For these women, there appears to be no benefit from using aspirin during pregnancy, while the risks from this drug to the newborn may be nontrivial. Unfortunately, the safety of alternative agents cannot be assumed. CONCLUSION

Data from the last 20 years establish that aspirin use at doses of a few aspirin tablets a day late in pregnancy results in hemostatic abnormalities in newborns. These abnormalities may include intracranial hemorrhage, particularly in preterm and low birth weight babies. Evidence is suggestive that exposure to analgesic/antipyretic doses (650-2,500 mg/day) late in pregnancy leads to increased maternal bleeding. Very low doses (< 81 mg/day) impaired platelet aggregation, but were not associated with overt bleeding. The length of labor and the length of gestation appear to be increased at doses of about 3,000 mg/day, but not at very low doses. There are no studies of this end point at typical analgesic/antipyretic doses. Epidemiologic studies show no aspirin-associated reduction in birth weight. Clinical and animal data suggest a possible effect on the pulmonary vasculature of the fetus, but no epidemiologic studies have been conducted on this end point.

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

While animal data show structural malformations to be induced by aspirin exposure, human studies have been less than convincing. In most instances, findings from one or two studies have not been followed up, and in the few instances where they have been, results have been inconsistent. In contrast to studies of effects from exposure late in pregnancy, studies of malformations are hampered by severe problems in assessing exposure early in pregnancy. With respect to beneficial effects, the evidence is strong that very low doses of aspirin (less than 100 mg/day) improve pregnancy outcome among women at high risk of pregnancy-induced hypertension, umbilical placental insufficiency, or repeat abortions associated with lupus anticoagulant. A full evaluation of risks at these low doses has not been conducted. While there are a great many gaps in knowledge, the degree of concordance between the findings from experimental animal studies and from epidemiologic investigations is striking. The data are clear for the responses mediated by inhibition of prostaglandin synthesis, particularly the platelet-related effects. Animal and human data are also consistent in showing prolongation of gestation, and there are suggestions of similarities in the findings related to constriction of the ductus arteriosus. Neither human nor animal data support an independent effect on birth weight. Where epidemiologic evidence and case series suggest teratogenicity, it is at some of the same sites at which animals have been sensitive. Both experimental and epidemiologic studies show some evidence for behavioral teratogenicity. The medicinal properties of salicin were discovered in the early part of the 1800s, and commercial production of acetylsalicylic acid under the name aspirin began in 1899. Aspirin has been used as an analgesic, antipyretic, and antiinflammatory agent for nearly a century. In the late 1970s, the Panel on Over-the-Counter Drugs of the US Food and Drug Administration concluded that women should be warned of the

ASPIRIN USE IN PREGNANCY

hazards associated with aspirin use in pregnancy (21). In 1988, the US Food and Drug Administration proposed the following warning label for products containing aspirin: "IMPORTANT: Do not take this product during the last three months of pregnancy unless directed by a doctor. Aspirin taken near the time of delivery may cause bleeding problems in both mother and child" (123). On April 6, 1990, the California Scientific Advisory Panel for Proposition 65, Safe Drinking Water and Toxic Enforcement Act of 1986, voted to list aspirin as a reproductive toxicant (133). Such actions are more than justified by the evidence. Whether there are other adverse effects of aspirin use in pregnancy, as suggested by human and animal studies, deserves closer investigation. REFERENCES

1. Heinonen OP, Slone D, Shapiro S. Birth defects and drugs in pregnancy. Littleton, MA: PSG Publishing Company, 1977. 2. Richards IDG. Congenital malformations and environmental influences in pregnancy. Br J Prev Soc Med 1969;23:218-25. 3. Nelson MM, Forfar JO. Associations between drugs administered during pregnancy and congenital abnormalities of the fetus. BMJ 1971; 1:523-7. 4. Forfar JO, Nelson MM. Epidemiology of drugs taken by pregnant women: drugs that may affect the fetus adversely. Clin Pharmacol Ther 1973; 14:632-42. 5. Bleyer WA, Au WYW, Lange WA, et al. Studies on the detection of adverse drug reactions in the newborn. I. Fetal exposure to maternal medication. JAMA 1970;213:2046-8. 6. Finnigan D, Burry AF, Smith ID. Analgesic consumption in an antenatal clinic survey. Med J Aust 1974;l:761-2. 7. Collins E, Turner G. Maternal effects of regular salicylate ingestion in pregnancy. Lancet 1975; 2:335-7. 8. Rothman KJ, Fyler DC, Goldblatt A, et al. Exogenous hormones and other drug exposures of children with congenital heart disease. Am J Epidemiol 1979; 109:433-9. 9. Streissguth AP, Treder RP, Ban- HM, et al. Aspirin and acetaminophen use by pregnant women and subsequent IQ and attention decrements. Teratology 1987;35:211-19. 10. Piper JM, Baum C, Kennedy DL. Prescription drug use before and during pregnancy in a Medicaid population. Am J Obstet Gynecol 1987; 157:148-56. 11. Palmisano PA, Cassady G. Salicylate exposure in the perinate. JAMA 1969;209:556-8. 12. Arrowsmith JB, Kennedy DL, Kuritsky JN, et

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

145

al. National patterns of aspirin use and Reye syndrome reporting, United States, 1980 to 1985. Pediatrics 1987;79:858-63. 13. Beaufils M, Uzan S, Donsimoni R, et al. Prevention of pre-eclampsia by early antiplatelet therapy. Lancet 1985;l:840-2. 14. Wallenburg HCS, Dekker GA, Makovitz JW, et al. Low-dose aspirin prevents pregnancy-induced hypertension and pre-eclampsia in angiotensinsensitive primigravidae. Lancet 1986;l:l-3. 15. 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 1989;321:357-62. 15a. Rothman KJ. No adjustments are needed for multiple comparisons. Epidemiology 1990;l:436. 16. Schiff E, Peleg E, Goldenberg M, et al. The use of aspirin to prevent pregnancy-induced hypertension and lower the ratio of thromboxane A2 to prostacyclin in relatively high risk pregnancies. N Engl J Med 1989;321:351-6. 17. Rudolph AM. Effects of aspirin and acetaminophen in pregnancy and in the newborn. Arch Intern Med 1981;141:358-63. 18. Collins E. Maternal and fetal effects of acetaminophen and salicylates in pregnancy. Obstet Gynecol 1981;58:57s-62s. 19. Niederhoff H, Zahradnik H-P. Analgesics during pregnancy. Am J Med 1983;75:117-2O. 20. Sibai BM, Amon EA. How safe is aspirin use during pregnancy? Contemp Obstet Gynecol 1988;32:73-84. 21. Corby DG. Aspirin in pregnancy: maternal and fetal effects. Pediatrics 1978;62:930-7. 22. Rainsford KD. Aspirin and the salicylates. London: Butterworth, 1984. 23. Flower RJ, Moncada S, Vane JR. Analgesicantipyretics and anti-inflammatory agents; drugs employed in the treatment of gout. In: Gilman AG, Goodman LS, Rail TW, et al., eds. Goodman and Gilman's the pharmacological basis of therapeutics. 7th ed. New York: MacMillan Publishing Company, 1985:674-715. 24. Kucera JL, Bullock FJ. The binding of salicylate to plasma protein from several animal species. J Pharm Pharmacol 1969;21:293-6. 25. Kimmel CA, Young JF. Correlating pharmacokinetics and teratogenic endpoints. Fundam Appl Toxicol 1983;3:250-5. 26. Wolff F, Berg R, Bolte A, et al. Perinatal pharmacokinetics of acetylsalicylic acid. Arch Gynecol 1982;233:15-22. 27. Klein KL, Clark KE, Scott WJ. Prostaglandin synthesis in rat embryo tissue: the effect of nonsteroidal anti-inflammatory drugs in vivo and ex vivo. Prostaglandins 1984;27:659-73. 28. Ylikorkala O, Makila UM, Kaapa P, et al. Maternal ingestion of acetylsalicylic acid inhibits fetal and neonatal prostacyclin and thromboxane in humans. Am J Obstet Gynecol 1986; 155:3459. 29. Tagashira E, Nakao K, Urbano T, et al. Salicylic acid in rats. Jpn J Pharmacol 1981;31:563-71. 30. Lovecchio JL, Krasner J, Yaffe SJ. Serum pro-

146

31.

32.

33.

34.

35.

36. 37.

38. 39.

40.

41. 42. 43. 44.

45. 46. 47. 48. 49.

HERTZ-PICCIOTTO ET AL. tein binding of salicylate during pregnancy and the puerperium. Dev Pharmacol Ther 1981;2: 172-9. Jankowski A, Skublick S, Wichlinski LM, et al. Clinical pharmcokinetic investigations of acetylsalicylic acid in cases of imminent premature delivery. J Clin Hosp Pharmacy 1985; 10:361-6. Levy G, Garrettson LK. Kinetics of salicylate elimination by newborn infants of mothers who ingested aspirin before delivery. Pediatrics 1974;53:201-10. Bleyer WA, Breckenridge RT. Studies on the detection of adverse drug reactions in the newborn. II. The effects of prenatal aspirin on newborn homeostasis. JAMA 1970;213:2049-53. Stuart MJ, Gross SJ, Elrad H, et al. Effects of acetylsalicylic acid ingestion on maternal and neonatal hemostasis. N Engl J Med 1982;307: 909-12. Zierler S, Rothman KJ. Congenital heart disease in relation to maternal use of bendectin and other drugs in early pregnancy. N Engl J Med 1985;313:347-52. Werler MM, Mitchell AA, Shapiro S. First trimester aspirin use in relation to congenital cardiac defects. N Engl J Med 1989,321:1639-42. Lewis RB, Schulman JD. Influence of acetylsalicylic acid, an inhibitor of prostaglandin synthesis, on the duration of human gestation and labor. Lancet 1973;2:1159-61. Mitchell AA, Cottier LB, Shapiro S. Effect of questionnaire design on recall of drug exposure in pregnancy. Am J Epidemiol 1986;123:670-6. Rumack CM, Guggenheim MA, Rumack BH, et al. Neonatal intracranial hemorrhage and maternal use of aspirin. Obstet Gynecol 1981;58:52s6s. Aiken JW. Aspirin and indomethacin prolong parturition in rats: evidence that prostaglandins contribute to expulsion of foetus. Nature 1972; 240:21-5. Jewell NP. On the bias of commonly used measures of association for 2 x 2 tables. Biometrics 1986;42:351-8. Guyton AC. Textbook of medical physiology. 6th ed. Philadelphia, PA: WB Saunders Company, 1981:67. Corby DG, Schulman I. The effects of antenatal drug administration on aggregation of platelets of newborn infants. J Pediatr 1971;79:307-13. Wolff F, Berg R, Bolte A. Clinical study of the labour inhibiting effects and side effects of acetylsalicylic acid (ASA). Geburtshilfe Frauenheilkd 1981;41:96-100. Schiff E, Mashiach S. Aspirin to prevent pregnancy-induced hypertension. (Letter). N Engl J Med 1990;322:204-5. Sibai BM, Mirro R, Chesney CM, et al. Lowdose aspirin in pregnancy. Obstet Gynecol 1989;74:551-7. MacFarlane MJ, Feinstein AR. Aspirin and maternal or neonatal hemostasis. (Letter). N Engl J Med 1983;308:281. Stuart MJ. Aspirin and maternal or neonatal hemostasis. (Letter). N Engl J Med 1983;308:281. Eriksson M. Salicylate induced foetal hemor-

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

50.

51.

52. 53. 54. 55.

56. 57. 58.

59.

60. 61.

62.

63. 64. 65. 66. 67.

rhage in two mouse strains. Acta Pathol Microbiol Scand 1969;76:164-70. Eriksson M. Salicylate induced foetal damage during late pregnancy in mice: the modifying effect of repeated administration and dosage. Acta Paediatr Scand 1970;59:517-22. Eriksson M. Salicylate-induced foetal damage during late pregnancy in mice: a comparison between sodium salicylate, acetylsalicylic acid and salicylsalicylic acid. Acta Pharmacol Toxicol 1971;29:250-5. Gulamhusein AP, Harrison-Sage C, Beck F, et al. Salicylate-induced teratogenesis in the ferret. Life Sci 1980;27:1799-1805. LaPointe R, Harvey EB. Salicylamide-induced anomalies in hamster embryos. J Exp Zool 1964;156:197-200. Eriksson M. Effect of salicylate treatment on fetal and maternal prothrombin time in the mouse. Acta Physiol Scand 1970;78:39-42. Casteels-Van Daele M, Jaeken J, Eggermont E, et al. More on the effects of antenatally administered aspirin on aggregation of platelets of neonates. (Letter). J Pediatr 1972;80:685-6. Haslam RR, Ekert H, Gillam GL. Hemorrhage in neonate possibly due to maternal ingestion of salicylate. J Pediatr 1974;84:556-7. Haslam RH. Neonatal purpura secondary to maternal salicylism. (Letter). J Pediatr 1975;86:653. Philip AGS, Allan WC, Tito AM, et al. Intraventricular hemorrhage in preterm infants: declining incidence in the 1980s. Pediatrics 1989;84:797801. National Center for Health Statistics, Public Health Service, US Department of Health and Human Services. Vital statistics of the United States 1986. Vol I. Natality. Washington, DC: US GPO, 1988. (Publication no. (PHS) 88-1123). van den Berg BJ, Oechsli FW. Prematurity. In: Bracken MB, ed. Perinatal epidemiology. New York: Oxford University Press, 1984:69-85. Behrman HR, Caldwell BV. Prostaglandins, thromboxanes and leukotrienes. In: Yen SSC, Jaffe RB, eds. Reproductive endocrinology: physiology, pathophysiology and clinical management. 2nd ed. Philadelphia, PA: WB Saunders Company, 1986:154-76. Chester R, Dukes M, Slater SR, et al. Delay of parturition in the rat by anti-inflammatory agents which inhibit the biosynthesis of prostaglandins. Nature 1972;240:37-8. Tuchmann-Duplessis H, Hiss D, Mottot G, et al. Effects of prenatal administration of acetylsalicylic acid in rats. Toxicology 1975;3:207-ll. O'Dell BL, Reynolds G, Reeves PG. Analogous effects of zinc deficiency and aspirin toxicity in the pregnant rat. J Nutr 1977;107:1222-8. Novy MJ, Cook MJ, Manaugh L. Indomethacin block of normal onset of labor in primates. Am J Obstet Gynecol 1974;118:412-16. TambyRaja RL, Salmon JA, Karim SMM, et al. F prostaglandin levels in amniotic fluid in premature labour. Prostaglandins 1977;13:339-48. Gyory G, Kiss C, Benyo R, et al. Inhibition of labour by prostaglandin antagonists in impending abortion and preterm and term labour. Lan-

ASPIRIN USE IN PREGNANCY cet 1974;2:293. 68. Reiss U, Atad J, Rubinstein I, et al. The effect of indomethacin in labour at term. Int J Gynaecol Obstet 1976;14:369-74. 69. Bakketeig LS, Hoffman HJ, Oakley ART. Perinatal mortality. In: Bracken MB, ed. Perinatal epidemiology. New York: Oxford University Press, 1984:99-151. 70. Butcher RE, Vorhees CV, Kimmel CA. Learning impairment from maternal salicylate treatment in rats. Nature New Biol 1972;236:211-12. 71. Khera KS. Teratogenicity studies with methotrexate, aminopterin, and acetylsalicylic acid in domestic cats. Teratology 1976;14:21-8. 72. Vorhees CV, Klein KL, Scott WJ. Aspirininduced psychoteratogenesis in rats as a function of embryonic age. Teratogenesis Carcinog Mutagen 1982;2:77-84. 73. Lubawy WC, Garrett RJB. Effects of aspirin and acetaminophen on fetal and placental growth in rats. J Pharm Sci 1977;66:111-12. 74. Turner G, Collins E. Fetal effects of regular salicylate ingestion in pregnancy. Lancet 1975;2: 338-9. 75. Shapiro S, Siskind V, Monson RR, et al. Perinatal mortality and birth-weight in relation to aspirin taken during pregnancy. Lancet 1976; 1:1375-6. 76. Trudinger BJ, Cook CM, Thompson RS, et al. Low-dose aspirin therapy improves fetal weight in umbilical placental insufficiency. Am J Obstet Gynecol 1988; 159:681-5. 77. Trudinger B, Cook CM, Thompson R, et al. Lowdose aspirin improves fetal weight in umbilical placental insufficiency. (Letter). Lancet 1988;2: 214-15. 78. Witter FR, Niebyl JR. Inhibition of arachidonic acid metabolism in the perinatal period: pharmacology, clinical application and potential adverse effects. Semin Perinatol 1986;10:316-33. 79. Peterson, RG. Consequences associated with nonnarcotic analgesics in the fetus and newborn. Federation Proc 1985;44:2309-13. 80. Levin DL, Fixler DE, Morriss FC, et al. Morphologic analysis of the pulmonary vascular bed in infants exposed in utero to prostaglandin synthetase inhibitors. J Pediatr 1978;92:478-83. 81. Levin DL. Effects of inhibition of prostaglandin synthesis on fetal development, oxygenation, and the fetal circulation. Semin Perinatol 1980;4:3544. 82. Momma K, Takeuchi H. Constriction of fetal ductus arteriosus by non-steroidal antiinflammatory drugs. Prostaglandins 1983;26: 631-43. 83. Takacs E, Warkany J. Experimental production of congenital cardiovascular malformations in rats by salicylate poisoning. Teratology 1968; 1:109-18. 84. Arcilla RA, Thilenius OG, Ranniger K. Congestive heart failure from suspected ductal closure in utero. J Pediatr 1969;75:74-8. 85. Warkany J, Takacs E. Experimental production of congenital malformations in rats by salicylate poisoning. Am J Pathol 1959;35:315-31. 86. Fritz H. The effect of cortisone on the teratogenic

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

147

action of acetylsalicylic acid and diphenylhydantoin in the mouse. Experientia 1976;32:721-2. 87. Tanigawa H, Obori R, Tanaka H, et al. Reproduction study of 4-ethoxy-2-methyl-5-morpholino-3(2H)-pyridazinone (M73101) in rats. II. Administration of M73101 during the period of major organogenesis. J Toxicol Sci 1979;4:175200. 88. Trasler DG. Aspirin-induced cleft lip and other malformations in mice. (Letter). Lancet 1965; 1:606-7. 89. Kimmel CA, Wilson JG, Schumacher HJ. Studies on metabolism and identification of the causative agent in aspirin teratogenesis in rats. Teratology 1971;4:15-24. 90. Kimmel CA, Butcher RE, Vorhees CV, et al. Metal-salt potentiation of salicylate-induced teratogenesis and behavioral changes in rats. Teratology 1974;10:293-300. 91. Robertson RT, Allen HL, Bokelman DL. Aspirin: teratogenic evaluation in the dog. Teratology 1979;20:313-20. 92. Wilson JG. Use of rhesus monkeys in teratological studies. Federation Proc 1971;30:104-9. 93. Beall JR, Klein MF. Enhancement of aspirininduced teratogenicity by food restriction in rats. Toxicol Appl Pharmacol 1977;39:489-95. 94. Wilson JG, Ritter EJ, Scott WJ, et al. Comparative distribution and embryotoxicity of acetylsalicylic acid in pregnant rats and rhesus monkeys. Toxicol Appl Pharmacol 1977;41:67-78. 95. McNeil JR. The possible teratogenic effect of salicylates on the developing fetus. Clin Pediatr 1973;12:347-50. 96. Klein KL, Scott WJ, Clark KE, et al. Indomethacin-placental transfer, cytotoxicity, and teratology in the rat. Am J Obstet Gynecol 1981; 141:448-52. 97. Slone D, Siskind V, Heinonen OP, et al. Aspirin and congenital malformations. Lancet 1976;1: 1373-5. 98. Winship KA, Cahal DA, Weber JCP, et al. Maternal drug histories and central nervous system anomalies. Arch Dis Child 1984;59:1052-60. 99. Thomas DC, Siemiatycki J, Dewar R, et al. The problem of multiple inference in studies designed to generate hypotheses. Am J Epidemiol 1985; 122:1080-95. 100. Crombie DL, Pinsent RJFN, Slater B, et al. Teratogenic drugs—RCGP survey. BMJ 1970;4:178-9. 101. Saxen I. Associations between oral clefts and drugs taken during pregnancy. Int J Epidemiol 1975;4:37-44. 102. Yokoyama A, Takakubo F, Eto K, et al. Teratogenicity of aspirin and its metabolite, salicylic acid, in cultured rat embryos. Res Commun Chem Pathol Pharmacol 1984;46:77-91. 103. Windholz M, Budavari S, Blumetti RF, et al., eds. The Merck index. An encyclopedia of chemicals, drugs, and biologicals. 10th ed. Rahway, NJ: Merck & Company, 1983:155-6. 104. Bellinger DC, Needleman HL, Leviton A, et al. Early sensory-motor development and prenatal exposure to lead. Neurobehav Toxicol Teratol 1984;6:387-402.

148

HERTZ-PICCIOTTO ET AL.

105. Dietrich KN, Krafft KM, Bier M, et al. Neurobehavioural effects of foetal lead exposure: the first year of life. In: Smith MA, Grant LD, Sors AI, eds. Lead exposure and child development, an international assessment. Boston, MA: Kluwer Academic Publishers, 1989:320-31. 106. Needleman HL, Rabinowitz M, Leviton A, et al. The relationship between prenatal exposure to lead and congenital anomalies. JAMA 1984; 251:2956-9. 107. Klebanoff MA, Berendes HW. Aspirin exposure during the first 20 weeks of gestation and IQ at four years of age. Teratology 1988;37:249-55. 108. Heymann MA, Rudolph AM. Effects of acetylsalicylic acid on the ductus arteriosus and circulation in fetal lambs in utero. Circ Res 1976; 38:418-22. 109. Gilman EA, Wilson LMK, Kneale GW, et al. Childhood cancers and their association with pregnancy drugs and illnesses. Paediatr Perinat Epidemiol 1989;3:66-94. 110. Walsh SW. Preeclampsia: an imbalance in placental prostacylin and thromboxane production. Am J Obstet Gynecol 1985;152:335-40 111. Fitzgerald DJ, Entman SS, Mulloy K, et al. Decreased prostacyclin biosynthesis preceding the clinical manifestation of pregnancy-induced hypertension. Circulation 1987;75:956-63. 112. Thorp JA, Walsh SW, Brath PC. Low-dose aspirin inhibits thromboxane, but not prostacyclin, production by human placental arteries. Am J Obstet Gynecol 1988;159:1381-4. 113. Sanchez-Ramos L, O'Sullivan MJ, GarridoCalderon J. Effect of low-dose aspirin on angiotensin II pressor response in human pregnancy. Am J Obstet Gynecol 1987;156:193-4. 114. Fitzgerald D, Fitzgerald G. Thromboxane production in placentas of women with preeclampsia. (Letter reply). Am J Obstet Gynecol 1990; 160:1536. 115. Walsh SW. Thromboxane production in placentas of women with preeclampsia. (Letter). Am J Obstet Gynecol 1990;160:1535. 116. Crandon AJ, Isherwood DM. Effect of aspirin on incidence of pre-esclampsia. (Letter). Lancet 1979;1:1356. 117. Wallenburg HCS, Rotmans N. Prevention of recurrent idiopathic fetal growth retardation by low-dose aspirin and dipyridamole. Am J Obstet Gynecol 1987;157:1230-5. 118. Feinstein DI. Lupus anticoagulant, thrombosis, and fetal loss. N Engl J Med 1985;313:1348-50. 119. Lubbe WF, Butler WS, Palmer SJ, et al. Fetal survival after prednisone suppression of maternal lupus-anticoagulant. Lancet 1983;l:1361-3. 120. Branch DW, Scott JR, Kochenour NK, et al. Obstetric complications associated with the lupus anticoagulant. N Engl J Med 1985;313:13226.

Downloaded from https://academic.oup.com/epirev/article-abstract/12/1/108/575174 by INSEAD user on 19 March 2018

121. Lubbe WF, Liggins GC. Lupus anticoagulant and pregnancy. Am J Obstet Gynecol 1985;153:3227. 122. Elder MG, de Swiet M, Robertson A, et al. Lowdose aspirin in pregnancy. (Letter). Lancet 1988;l:410. 123. US Department of Health and Human Services, Food and Drug Administration. Internal analgesic, antipyretic, and antirheumatic drug products for over-the-counter human use: tentative final monograph; notice of proposed rulemaking. Federal Register 46256. Vol 53, no. 221, 21 CFR parts 310, 343, and 369. Washington, DC: US GPO, 1988. 124. Larsson KS, Bostrom H, Ericson B. Salicylateinduced malformations in mouse embryos. Acta Paediatr (Stockh) 1963;52:36-40. 125. Koshakji RP, Schulert AR. Biochemical mechanisms of salicylate teratology in the rat. Biochem Pharmacol 1973;22:407-16. 126. Needleman HL. The persistent threat of lead: a singular opportunity. Am J Public Health 1989;79:643-5. 127. Eskenazi B. Behavioral teratology. In: Bracken MB, ed. Perinatal epidemiology. New York: Oxford University Press, 1984:216-54. 128. Olsen SF, Hansen HS, Sorensen TIA, et al. Intake of marine fat, rich in (n-3)-polyunsaturated fatty acids, may increase birthweight by prolonging gestation. Lancet 1986;2:367-9. 129. Andersen HJ, Andersen LF. Diet, pre-eclampsia, and intrauterine growth retardation. (Letter). Lancet 1989;1:1146. 130. Spitz B, Magness RR, Cox SM, et al. Low-dose aspirin. I. Effect on angiotensin II pressor responses and blood prostaglandin concentrations in pregnant women sensitive to angiotensin II. Am J Obstet Gynecol 1988;159:1035-43. 131. Ferrazzani S, Caruso A, De Carolis S, et al. Individualized chronobiologic assessment of lowdose aspirin for pre-eclampsia prevention. (Proceedings of the 19th International Conference, International Society for Chronobiology, Bethesda, MD, June 20-24, 1989). Chronobiologia 1989;16:134. 132. De Carolis S, Caruso A, Ferrazzani S, et al. Lowdose aspirin, the angiotensin sensitivity test (AST) and blood pressure in pregnancy: chronobiologic group assessment. (Proceedings of the 19th International Conference, International Society for Chronobiology, Bethesda, MD, June 2024,1989). Chronobiologia 1989;16:124. 133. California Scientific Advisory Panel. Discussion of aspirin as a reproductive toxicant. Meeting of the California Scientific Advisory Panel for Proposition 65 (Safe Drinking Water and Toxic Enforcement Act of 1986). Wyatt Pavilion, University of California, Davis, CA, April 6,1990.

The risks and benefits of taking aspirin during pregnancy.

EPIDEMIOLOGIC REVIEWS Copyright © 1990 by The Johns Hopkins University School of Hygiene and Public Health Allrightsreserved Vol. 12, 1990 Printed i...
3MB Sizes 0 Downloads 0 Views