Serjeant, Graham R. MD, FRCP*; Loy, Luana Look MB*; Crowther, Mark MB*; Hambleton, Ian R. PhD†; Thame, Minerva DM(Paed), PhD‡
Early reports on the outcome of pregnancy in homozygous sickle cell (SS) disease depicted an almost universal adverse outcome for mother and child. This outcome has greatly improved,1 but there remain potentially serious biases in existing reports because the patients have been ascertained usually through presentation at antenatal clinics, implying that consistent early pregnancy losses are underestimated, and also because of the biases inevitable in hospital-based series. We hypothesize that these factors will have distorted aspects of pregnancy outcome in sickle cell disease. Removal of such biases is difficult, and the most representative data available are from the cooperative study in the United States2 of 320 pregnancies in 155 women with homozygous sickle cell disease recruited from 19 centers from 1979 to 1986. Such multicenter studies, although carefully conducted and controlled, inevitably introduce confounding factors such as variability in management between centers. A Jamaican retrospective study3 reporting 664 pregnancies among 297 women with homozygous sickle cell disease over the period 1959–1984 had the advantages of a single center but the disadvantage of secular and management change over a 25-year period. We believe that the best opportunity for describing true pregnancy outcome is in a representative sample of patients in a cohort study from birth. Such a cohort also has the advantages of documenting age at menarche and interval to first pregnancy and of recording recurrent early pregnancy losses. The Jamaican Cohort Study, in which women with homozygous sickle cell disease (subjects) and women with a normal hemoglobin genotype (controls) have been followed from birth, now provides that opportunity.
SUBJECTS AND METHODS
A cohort study from birth,4,5 based on the screening of 100,000 consecutive deliveries at the main government maternity hospital (Victoria Jubilee Hospital, Kingston, Jamaica) from June 1973, detected 315 babies with homozygous sickle cell disease over an 8-year period. The first 125 babies were each matched with 2 controls with a normal hemoglobin (AA) phenotype of the same sex, born closest in time (usually immediately before and after). This resulted in 250 controls collected over the first 2 years. The subsequent progress of these subjects and their recruitment to the present study is shown in Figure 1. At the study date (May 31, 2003), the longer recruitment for mothers with homozygous sickle cell disease resulted in a wider age range (21.6–29.8 years) than in AA controls (27.4–29.9).
Subjects and controls were followed at 3- to 6-month intervals and encouraged to attend at any time when sick, and defaulters were actively traced, with rates of annual default (missed appointments for 1 year) consistently below 5% in subjects with homozygous sickle cell disease. Clinical management was generally supportive and comprehensive but without specific interventions such as chronic transfusion or hydroxyurea. Hematology was monitored at 6-month intervals when the subjects and controls were well, routine hematological indices measured in electronic counters (Coulter Electronics, Miami, FL) and fetal hemoglobin by alkali denaturation.6 Subjects were routinely questioned about suspected or confirmed pregnancies, and 93% of subjects and controls were systematically reviewed in the 3 weeks before the study date. Except for 3 deliveries (2 controls, 1 subject) for which the notes were untraceable, the hospital notes of all deliveries were reviewed. Prophylactic transfusion in the management of pregnancy is not performed in Jamaica, but symptomatically indicated transfusions were given during 10 pregnancies in 9 subjects (4 acute chest syndrome, 3 postpartum hemorrhage, 2 anemia, 1 multiple complications) and to 4 controls (2 postpartum hemorrhage, 2 anemia). Some proceeded to delivery without any antenatal care (9 [18%] subjects, 47 [41%] controls) although median antenatal attendance was 6 visits (range 0–18) for subjects and 3 visits (range 0–16) for controls.
Spontaneous pregnancy loss was defined as an abortion before 24 weeks of gestation and as a stillbirth after this age. The diagnosis of abortion was based on history and not histologically confirmed. In all early losses, pregnancy was confirmed by positive pregnancy tests. The term neonatal death was applied to children who were born alive but died within 28 days. Gestational age was calculated from last menstrual period (LMP) and was considered accurate in 152 subjects (43 subjects, 109 controls). Age at first pregnancy was defined by the outcome date to avoid inaccuracies in recalled LMP. Prematurity or preterm delivery referred to a gestational age less than 259 days (37 weeks) and low birth weight as below 2,500 g. Because ultrasound examinations were not routinely performed, statements about the frequency of intrauterine growth restriction were not possible. Acute chest syndrome was defined as pulmonary symptoms and signs associated with a new pulmonary infiltrate on chest radiograph. Painful crisis was typical sickle-related bone pain of sufficient severity to require narcotic analgesia. Pregnancy-induced hypertension was defined as a systolic pressure of 140 mm Hg or greater or a diastolic pressure of 90 mm Hg or greater after 20 weeks of gestation in the absence of proteinuria in mothers who were not chronically hypertensive and in whom blood pressure returned to normal by 1 month after delivery.7 Preeclampsia was defined by the same blood pressure criteria but with proteinuria. Proteinuria was defined by Dipstick as trace proteinuria or more in subjects and 1+ or more in controls. The difference in definitions recognizes the less concentrated urine in homozygous sickle cell disease; lesser amounts were considered as absence of proteinuria. Eclampsia referred to preeclampsia and convulsions in a subject without prior neurologic disorders. Urinary tract infection applied to urinary symptoms with positive urine culture. Antepartum hemorrhage was considered to be any vaginal bleeding before labor, and postpartum hemorrhage was an estimated loss of 500 mL or more during or following delivery or loss of 1,000 mL or more with cesarean delivery. A retained placenta was defined as retention for 60 minutes or more after delivery of the infant. Hospital admissions were categorized as antepartum admissions (for problems preceding the onset of labor), intrapartum admissions, and postpartum admissions. Default referred to Sickle Cell Clinic attendance and was used for intervals of missed scheduled clinic attendance; periods of default might span several appointments and sometimes years.
Age at first pregnancy and the interval between menarche and first pregnancy were calculated by Kaplan-Meier estimates, and differences were tested by the log-rank test. Because subjects were recruited over a longer period and younger subjects had less time to become pregnant, survival calculations also controlled for year of birth using the Cox proportional hazards model. The probability of having a baby (yes/no) was compared in subjects and controls using logistic regression, and the number of births was compared using linear regression. Pregnancy outcome and pregnancy-related complications were assessed by the t test for mean summaries, the χ2 test for counts of events, and Fisher exact test when counts included values less than 10. Because 15 univariate tests were performed, there was a reasonable chance of a false positive result, and a family-wise error rate correction8 required a P value of .003 before assuming significance at the 5% level. Potential determinants of birth weight were assessed with multiple linear regression by examining the predictive effects of gestational age (weeks), mother's age (years), parity, mother's prepregnancy weight, and babies’ gender. Three regression analyses were performed, in both genotypes combined and in controls and subjects separately. The last regression in the SS subjects also considered the potential predictive effects of maternal total hemoglobin and fetal hemoglobin levels, defined as the average of all steady state observations over the age of 5 years. The regression also considered the total number of clinical events (painful crisis, acute chest syndrome, and urinary tract infection) during pregnancy, up to 1 week before delivery, as a single continuous variable. A robust estimate of standard error9 was used to account for multiple births by the same mother in all univariate tests and regressions. All analyses were conducted using commercially available software (Stata 8, Stata Statistical Software, College Station, TX).
The cohort study was approved by the Ethical Committee of the University Hospital of the West Indies. Mothers whose children were recruited to the cohort gave informed consent, and each subject was reinterviewed at age 18 years and gave informed consent to continue in the study. Specific permission was obtained to present pregnancy data.
The structure of the study group (Figure 1) shows 94 completed eligible singleton pregnancies in 52 subjects and 157 pregnancies in 68 controls. The only twin pregnancy, in a multiparous control, was excluded because twin pregnancies change many of the features associated with pregnancy outcome. The mean number of pregnancies in those becoming pregnant was 1.8 (range 1–5) in subjects and 2.3 (range 1–6) among controls.
Median age at menarche was 15.4 years (standard deviation 1.6 years) in subjects and 13.0 years (standard deviation 1.1 year) in controls for a significant mean delay of 2.38 years and median delay of 2.45 years, similar to previous reports in this group.10 First pregnancy occurred earlier in the controls (median age 20.1 years, 95% confidence interval [CI] 19.2, 21.3 versus 23.7 years, 95% CI 21.5, 26.5; hazard ratio 1.8, 95% CI 1.2, 2.5, P = .003) (Figure 2), but the difference was largely accounted for by the different age structure of the groups because, after adjustment for year of birth, the difference was only marginally significant (hazard ratio 1.6, 95% CI 1.0, 2.4, P = .04). The interval between menarche and first pregnancy was slightly shorter in controls (median interval 7.1 years, 95% CI 6.2, 8.9 versus 9.0 years, 95% CI 6.8, 10.5), but this difference did not reach significance before or after adjustment for year of birth (adjusted hazard ratio 1.1, 95% CI 0.7, 1.8, P = .58). A lower proportion of subjects became pregnant; logistic regression showed controls were more likely to become pregnant (odds ratio 2.4 times, 95% CI 1.3, 4.4, P = .004), but this was due to the fact that controls were older because, after adjustment for year of birth, the difference was no longer significant (odds ratio of controls versus subjects 1.2, 95% CI 0.5, 3.0, P = .62). The number of pregnancies per pregnant patient suggested a greater pregnancy rate among controls (odds ratio 0.9 more pregnancies, 95% CI 0.3, 1.4, P = .004), but this was reduced to nonsignificant levels after adjustment for year of birth (odds ratio 0.5, 95% CI –0.3, 1.2, P = .24).
Compared with controls, subjects had significantly more spontaneous abortions, fewer live deliveries, a lower mean birth weight, a lower mean gestational age at delivery, and a greater proportion of premature infants (Table 1). The frequency of termination of pregnancy was similar between genotypes, and termination generally was for social reasons, although 2 controls required terminations for ectopic pregnancies. In 12 subjects, all pregnancies ceased as abortions (6 subjects, single abortion; 3, two abortions; 2, three abortions; 1, five abortions). This group accounted for 23 (77%) of the abortions. Furthermore, abortions tended to occur early, 24 (80%) occurring before 12 weeks. There were 6 stillbirths in subjects: 2 mothers with uncomplicated clinical courses, 1 mother with sickle-related admissions, 2 mothers with severe preeclampsia 2 and 6 days before fetal losses, and 1 infant with the cord twisted tightly around the neck. Successful pregnancy outcomes occurred in 57% of subjects (48 live deliveries among 33 subjects) and in 89% of controls (128 live deliveries among 63 controls).
The relationship between birth weight and gestational age was analyzed in 43 live births among 33 subjects and 109 live births among 62 controls after exclusion of pregnancies without reliable estimates of gestational age (5 subjects, 17 controls) or birth weight (2 controls). There was some evidence of a deceleration in birth weight increase as gestational age increased, but the effect was not significantly better than a linear summary (deviance difference = 2.1, P = .59), and a linear effect of age (Figure 3) was assumed in all regressions. Regression analysis of possible predictors of birth weight confirmed an effect of maternal genotype and showed influences of gestational age and maternal weight within each genotype (Table 2) and of maternal age and the baby's gender in controls but not subjects. There was a striking effect of antepartum clinical events in subjects, shown by restricting analysis to 3 events (13 painful crises, 5 acute chest syndrome, 8 urinary tract infections) as a single categorical variable with 3 levels (0, 1, 2). A single event lowered the birth weight by 0.35 kg (95% CI –0.65, –0.04, P = .03) and 2 events by 0.49 kg (95% CI –0.89, –0.09, P = .02). Further analysis indicated that this effect was due to painful crisis and urinary tract infections, with no contribution from acute chest syndrome. Inclusion of the number of clinical events in the regression model increased the variance in birth weight accounted for from 53% to 67%.
In our analysis of sequential pregnancy performance, there was no clear trend in the outcome of successive pregnancies.
Mode of delivery did not differ between genotypes, and there were no differences in indications for cesarean delivery, which included failure to progress (2 subjects, 4 controls), fetal distress or jeopardy (2 subjects), disproportion (3 controls), and preeclampsia or eclampsia (1 subject, 3 controls).
Among 54 pregnancies (48 live, 6 stillbirth) in 35 subjects, there were 14 painful crises in 11 (20%) pregnancies, accounting for 13 antepartum admissions in 10 subjects, mostly in the third trimester. There were 12 episodes of acute chest syndrome in 11 (20%) pregnancies, accounting for 5 antepartum admissions and 5 postpartum admissions. Urinary tract infections occurred in 8 (15%) pregnancies. Among 129 pregnancies (128 live, 1 stillbirth) in 63 controls, there were 12 antepartum admissions in 8 subjects, mostly for nonspecific abdominal pain. There were no genotype differences in pregnancy-induced hypertension (9 subject versus 8 control pregnancies; Fisher exact test, P = .20), preeclampsia (8 versus 7; Fisher exact test, P = .27), antepartum hemorrhage (1 versus 3; Fisher exact test, P = .99), or postpartum hemorrhage (4 versus 7; Fisher exact test, P = .99). Eclampsia occurred in a single pregnancy of each genotype. An unusually adherent or retained placenta occurred in 5 pregnancies of subjects, 2 retained for 140 and 180 minutes and one for 52 minutes, which required immediate manual removal because of an associated postpartum hemorrhage of 1,600 mL. The other 2 cases of retained placenta were associated with cesarean deliveries, one requiring curettage for retained products 10 days after section and the other noting an abnormally adherent placenta at the time of delivery. These 5 events suggested that this complication is marginally more common in homozygous sickle cell disease (Fisher exact test, P = .007, after adjustment for multiple testing, which required P < .003 for significance at the 5% level).
There were 2 deaths, both among subjects, for a mortality rate of 2.1% (95% CI 0%, 5.2%). An 18.5-year-old primipara from a rural area presented in labor at 35 weeks of gestation, having received no antenatal care. Urine testing showed 3+ proteinuria and 3+ blood, blood pressure 150/100 mm Hg, hemoglobin 3.7 g, and platelets 89,000. She delivered a live male child weighing 2,950 g and died from uncontrolled postpartum hemorrhage, autopsy findings being consistent with disseminated intravascular coagulation. The other death was a 17.2-year-old primipara with an uneventful pregnancy and delivery, but 6 weeks later she presented with an upper respiratory tract infection and 13 weeks postpartum with right lower lobe consolidation. She improved on therapy, but 1 week later became acutely ill and died, autopsy showing thrombosis of the right popliteal vein and bilateral pulmonary thromboembolism. The occurrence of deep vein thrombosis within 14 weeks of pregnancy is consistent with a pregnancy-related complication.
The cohort nature of the present study has provided new information in several areas. The delay in onset of menarche in homozygous sickle cell disease is well recognized, and the median delay of 2.45 years is similar to other reports, but the similar interval between menarche and first pregnancy in subjects and controls (data not previously available) conflicts with the common assumption of a relative infertility or lack of sexual exposure in homozygous sickle cell disease. Once pregnancy occurred, only 57% of subject pregnancies had a live outcome, compared with 89% of controls, and the major source of fetal loss was spontaneous abortion.
Observations that 77% of abortions occurred in women who had never had live deliveries and that 80% of abortions occurred before 12 weeks imply that many of these subjects would not have presented to obstetricians, and under normal circumstances, these abortions might have passed unrecorded. Spontaneous abortion rates were 19–24% in recent reports,11 an exception being the 5.9% rate from the Cooperative Study,2 which will have been affected by the 29.4% rate of elective abortions. Even within previous Jamaican studies, abortion rates have varied from 9.4% to 13.3%,3,12,13 figures that did not differ from estimates in the normal population. The abortion rate of 36% is much higher than in previous series but is believed to be more accurate because detailed questioning about menstrual irregularities and the widespread use of home testing pregnancy kits has confirmed pregnancies which might otherwise have been overlooked. Similar questioning of the normal controls produced an abortion rate of 10%, which is also believed to be a more reliable estimate. The stillbirth rate among subjects was not significantly higher, after adjustment for multiple testing.
In the present study, the mean gestational age was significantly lower than in controls (37.0 versus 38.7 weeks), but there was no evidence for secular change and no clear hematological or clinical associations with gestational age. This is consistent with reports in the literature since 1972, which have varied from 34.1 weeks14 to 38.5 weeks for deliveries,1 with some evidence of a secular increase in gestational age attributed to improved perinatal care. Preterm deliveries in our study occurred in 44%, higher than in previous reports (20–33%) and possibly influenced by earlier operative intervention in Jamaica, because the cesarean deliveries occurred at a significantly earlier age in SS mothers. The proportion of babies with low birth weights (42%) among subjects was similar to the 31–55% of previous reports, but the difference did not reach significance because of the adjustment needed for multiple testing. Even the controls had a greater frequency of low birth weight babies (19%), compared with estimates of 11.8% born to African-American mothers.15 The present study confirmed that the gestational age, maternal genotype, and maternal weight were predictors of birth weight in both genotypes and that 2 hematological indices (total hemoglobin and fetal hemoglobin) often influencing other clinical complications had no effect as determinants of birth weight. The effect on birth weight of clinical events during pregnancy in subjects was remarkable considering the low frequency of events, and although demonstrable with painful crisis and urinary tract infection, the lack of an effect of acute chest syndrome may have resulted from only 5 antepartum events. This observation suggests that better maternal health during pregnancy should result in increasing birth weight and conflicts with the results of transfusion therapy in pregnancy16 which, despite reducing maternal painful crises and acute chest syndrome, had no effect on birth weight. Further analysis of the maternal complications in pregnancy among women with homozygous sickle cell disease and their effect on birth weight may give insights into the mechanisms involved.
Gestational hypertension has been reported to be more common in homozygous sickle cell disease,17 but there was no difference in pregnancy-induced hypertension or preeclampsia in the present study, even though the definition of preeclampsia was expanded for homozygous sickle cell disease as described in the text. Antepartum or postpartum hemorrhage did not differ between genotypes although a retained or unusually adherent placenta appeared marginally more common in subjects, which has not been previously reported. The shearing forces associated with contraction of the uterine myometrium is believed to be responsible for placental detachment, and although uterine atony has been reported in 10% of one series,1 this syndrome was not defined and there was no mention of placental retention. The tendency to placental retention or abnormal adherence raises the possibility that a sickle-induced scarring binds the placenta more firmly to the underlying uterine wall.
Two subjects died from pregnancy-associated complications, for a mortality rate of 2.1%, which is between 100 and 500 times greater than estimates in the general population, but one death with no antenatal monitoring might have been avoided with better care. The other death occurred 14 weeks after delivery and, with the short time course of massive pulmonary embolism, may have been unavoidable. Maternal mortality rates were similar in the United States until the early 1970s,1,18 but major series over the last 25 years have had no maternal mortality.11,14,19–21 In Africa, mortality rates of 7–12% persist,22–25 reflecting limited services and lack of antenatal care. The mortality rate in the present study sounds daunting, but based on only 2 deaths, the confidence intervals for this estimate are wide.
The outcome of pregnancy in homozygous sickle cell disease remains variable and unpredictable. Pregnancies were entirely uneventful in 24%, and even within an African environment where obstetric outcome is frequently poor, 28% of pregnancies in sickle cell disease in Uganda occurred without complications.24 The poor compliance with antenatal care, absent in 18% of subjects and 41% of controls, is to be deplored and represents an area where increased patient education could benefit pregnancy outcome.1 The spontaneous abortion rate of 36% remains a major contributor to fetal loss in homozygous sickle cell disease, and a better understanding of the mechanisms involved could make a major contribution to the successful outcome of pregnancy.
1. Powars DR, Sandhu M, Niland-Weiss J, Johnson C, Bruce S, Manning PR. Pregnancy in sickle cell disease. Obstet Gynecol 1986;67:217–28.
2. Smith JA, Espeland M, Bellevue R, Bonds D, Brown AK, Koshy M. Pregnancy in sickle cell disease: experience of the Cooperative Study of Sickle Cell Disease. Obstet Gynecol 1996;87:199–204.
3. Poddar D, Maude GH, Plant MJ, Scorer H, Serjeant GR. Pregnancy in Jamaican women with homozygous sickle cell disease: fetal and maternal outcome. Br J Obstet Gynaecol 1986;93:727–32.
4. Serjeant BE, Forbes M, Williams LL, Serjeant GR. Screening cord bloods for detection of sickle cell disease. Clin Chem 1974;20:666–9.
5. Serjeant GR, Serjeant BE, Forbes M, Hayes RJ, Higgs DR, Lehmann H. Haemoglobin gene frequencies in the Jamaican population: a study of 100,000 newborns. Br J Haematol 1986;64:253–62.
6. Betke K, Marti HR, Schlicht I. Estimation of small percentages of foetal haemoglobin. Nature 1959;184:1877–8.
7. Report of the National High Blood Pressure Education Program Working Group on High Blood Pressure in Pregnancy. Am J Obstet Gynecol 2000;183:S1–22.
8. Sidak Z. Rectangular confidence regions for the means of multivariate normal distributions. J Am Stat Assoc 1967;62:626–33.
9. White H. Maximum likelihood estimation of misspecified models. Econometrica 1982;50:1–25.
10. Serjeant GR, Singhal A, Hambleton IR. Sickle cell disease and age at menarche in Jamaican girls: observations from a cohort study. Arch Dis Child 2001;85:375–8.
11. Milner PF, Jones BR, Dobler J. Outcome of pregnancy in sickle cell anemia and sickle cell-hemoglobin C disease. Am J Obstet Gynecol 1980;138:239–45.
12. Anderson MF. The foetal risks in sickle cell anaemia. West Indian Med J 1971;20:288–95.
13. Morris JS, Dunn DT, Poddar D, Serjeant GR. Haematological risk factors for pregnancy outcome in Jamaican women with homozygous sickle cell disease. Br J Obstet Gynaecol 1994;101:770–3.
14. Sun PM, Wilburn W, Raynor BD, Jamieson D. Sickle cell disease in pregnancy: twenty years of experience at Grady Memorial Hospital, Atlanta, Georgia. Am J Obstet Gynecol 2001;184:1127–30.
15. Kleinman JC, Kessel SS. Racial differences in low birth weight: trends and risk factors. N Engl J Med 1987;317:749–53.
16. Koshy M, Burd L, Wallace D, Moawad A, Baron J. Prophylactic red-cell transfusion in pregnant patients with sickle cell disease: a randomized cooperative study. N Engl J Med 1988;319:1447–52.
17. National Heart, Lung, and Blood Institute, Division of Blood Diseases and Resources. Contraception and pregnancy. In: The management of sickle cell disease. 4th
ed. NIH Publication 02–2117. Bethesda (MD): National Institutes of Health; 2002. p. 145–51.
18. Fort AT, Morrison JC, Barreras L, Diggs LW, Fish SA. Counseling the patient with sickle cell disease about reproduction: pregnancy outcome does not justify the maternal risk! Am J Obstet Gynecol 1971;111:324–7.
19. Freeman MG, Ruth GJ. SS disease, SC disease, and CC disease: obstetric considerations and treatment. Clin Obstet Gynecol 1969;12:134–56.
20. Pritchard JA, Scott DE, Whalley PJ, Cunningham FG, Mason RA. The effects of maternal sickle cell hemoglobinopathies and sickle cell trait on reproductive performance. Am J Obstet Gynecol 1973;117:662–70.
21. Tuck SM, Studd JWW, White JM. Pregnancy outcome in sickle cell disease in the U.K. Br J Obstet Gynaecol 1983;90:112–17.
22. Hendrickse JP, Harrison KA, Watson-Williams EJ, Luzzatto L, Ajabor LN. Pregnancy in homozygous sickle-cell anaemia. J Obstet Gynaecol Br Commonw 1972;79:396–409.
23. Odum CU, Anorlu RI, Dim SI, Oyekan TO. Pregnancy outcome in HbSS-sickle cell disease in Lagos, Nigeria. West Afr J Med 2002;21:19–23.
24. Ndugwa CM. Pregnancy in sickle cell anaemia in Uganda (1971–1980). East Afr Med J 1982;59:320–6.
25. Dare FO, Makinde OO, Faasuba OB. The obstetric performance of sickle cell disease patients and homozygous hemoglobin C patients in Ile-Ife, Nigeria. Int J Gynaecol Obstet 1992;37:163–8.