Liver transplant recipients and their infants may have an increased risk of obstetric complications. Our objective was to describe pregnancy outcomes in women with a prior transplant from a population-based perspective. We analyzed the 1993–2005 US Nationwide Inpatient Sample database to identify obstetric hospitalizations among transplant recipients (n = 206) and controls matched by age, hospital, and year (n = 4060). The effect of prior transplantation on maternal and fetal outcomes was evaluated with regression models with adjustments for patient and hospital factors, including admission to a transplant center. Between 1993 and 2005, 146 delivery admissions among liver transplant recipients were identified. Cesarean deliveries were more common among transplant recipients (38% versus 24%; P = 0.0001); however, this difference was not significant after multivariate adjustment [OR (odds ratio) = 0.87; 95% confidence interval (CI) = 0.60–1.27]. Maternal mortality was similar among cases and controls (0% versus 0.02%; P = 1.00), but transplant patients had higher rates of fetal mortality (6.3% versus 2.0%; P = 0.0006), antepartum admission (OR = 2.27; 95% CI = 1.59–3.25), and maternal (OR = 2.63; 95% CI = 1.82–3.80) and fetal complications (OR = 2.49; 95% CI = 1.68–3.70). Gestational hypertension (30% versus 9%; P < 0.0001) and postpartum hemorrhage (8% versus 3%; P = 0.009) were more common among transplant recipients; their infants had higher rates of prematurity (27% versus 11%; P < 0.0001), distress (10% versus 5%; P = 0.005), and growth restriction (5% versus 2%; P = 0.05) but not congenital anomalies. Hospitalization in a transplant center (∼50%) was associated with similar obstetric outcomes. In conclusion, although most pregnancy outcomes are favorable, liver transplant recipients and their infants have an increased risk of obstetric complications. Additional studies evaluating mechanisms aimed at reducing these complications are necessary. Liver Transpl 16:56–63, 2010. © 2009 AASLD.
In the United States, there are more than 3000 female liver transplantation (LT) recipients of childbearing age.1 Since the first reported pregnancy in an LT patient in 1978,2 hepatologist and obstetric experience with managing pregnancy in this population has increased substantially. The literature suggests that most pregnancies in LT recipients have successful outcomes for the mother and child. However, an increased risk of complications, including prematurity, low birth weight, hypertension, renal dysfunction,3 and cesarean delivery, has been reported.4–7 These data have largely been derived from case reports,2, 8 registries [eg, the National Transplantation Pregnancy Registry (NTPR)9 and the UK Transplant Pregnancy Registry10], and single-center studies.3, 11–16 Although these reports have provided valuable information for counseling pregnant transplant patients, they are limited in some circumstances by small sample sizes and various biases, including reporting bias in voluntary registries and selection bias in studies from referral centers.
In light of these limitations, using a US population-based, administrative database, we examined obstetric outcomes among LT recipients. Our objective was to determine the impact of prior transplantation on maternal and fetal outcomes and to determine independent predictors of complications, including hospitalization in LT centers. We hypothesized that patients managed in transplant centers would have improved outcomes because of greater physician experience and multidisciplinary management.
PATIENTS AND METHODS
Data were extracted from the Healthcare Cost and Utilization Project Nationwide Inpatient Sample (NIS) database from 1993 to 2005.17 The NIS is the largest all-payer database of national hospital discharges (∼8 million/year) maintained by the Agency for Healthcare Research and Quality. It represents a 20% stratified random sample of nonfederal acute-care hospitals in the United States, including community, general, and academic centers. Stratified random sampling ensures that the database is representative of the US population and accounts for approximately 90% of all hospitalizations. Each data entry includes demographic details, diagnoses, procedures, admission type, hospital charges, length of stay (LOS), and hospital characteristics. Because each record is for a single hospitalization, not a person, there could be multiple records for individuals with repeat hospitalizations. NIS data compare favorably with the National Hospital Discharge Survey, and this supports its validity.18
We used International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis and procedure codes to identify pregnancy-related hospitalizations.19 Delivery admissions included a code for delivery, and antepartum admissions included a pregnancy code without a delivery code (see the appendix for specific ICD-9-CM codes). Among these hospitalizations, prior recipients of LT (V42.7 and 996.82) were matched with up to 20 randomly selected controls on the basis of age (±2 years), hospital, and calendar year. Over 99% of cases were matched with 20 controls. Patients who underwent transplantation (procedure code 50.5) during an obstetric admission were excluded.
Outcome Measures and Predictor Variables
We identified the following obstetric complications (see the appendix for codes): cesarean delivery, preterm labor (<37 weeks), premature rupture of membranes (<24 hours before labor), and placental abruption and previa. Fetal complications included death (including spontaneous and missed abortions), growth restriction, fetal distress, asphyxia, and congenital anomalies. Maternal outcomes included death, gestational diabetes, hypertension complicating pregnancy, antepartum and postpartum hemorrhage, blood transfusion, puerperal infection, and venous thromboembolism. We also examined LOS and hospital charges adjusted for inflation to 2005 dollars using the US Consumer Price Index.20
Our primary exposure variable was prior transplantation. Covariates included age, sex, race, insurance, admission status (elective versus urgent/emergent and transferred versus not transferred), and year (1993–1999 versus 2000–2005). For case-mix adjustment, we used the Elixhauser list of comorbidities, a well-validated algorithm for predicting hospital outcomes (categorized as 0, 1, or ≥2).21, 22 One of these comorbidities—hypertension—was not included in our risk adjustment models because the majority (93%) of patients with gestational hypertension (an outcome of interest) also had codes for preexisting hypertension due to overlap in coding for these conditions. Hospital-level covariates included the geographic region and whether or not the hospital was an LT center, which was defined as any hospital that performed at least one LT in a given year.
Data were analyzed with SAS version 9.1.3 (SAS Institute, Carey, NC) and SAS-callable SUDAAN version 9.0.1 (Research Triangle Institute, Research Triangle Park, NC). Discharge-level weights published by the Healthcare Cost and Utilization Project were used to estimate the number of obstetric hospitalizations in the United States during the study period.23 We compared patient characteristics and outcomes with Fisher's exact, chi-square, and Mann-Whitney tests and logistic regression models. Multivariate logistic regression models examining predictors of outcomes included adjustments for patient demographics (age, sex, race, and insurance), comorbidities, admission characteristics, year, region, and whether or not the hospital was an LT center. Multiple linear regression analyses were also used to adjust for confounders in comparing LOS and hospital charges, which were logarithmically transformed because of their skewed distributions. A P value less than 0.05 was considered statistically significant.
Between 1993 and 2005, there were 206 obstetric hospitalizations among women who had undergone LT previously in the NIS database. Sixty (29%) of these admissions were antepartum, and 146 (71%) included the delivery of 148 infants. These delivery admissions correspond to approximately 736 deliveries in the United States during the study period. The estimated average annual number of deliveries in post-LT patients increased from 47 during 1993–1999 to 68 during 2000–2005. There were several differences in the baseline characteristics of obstetric admissions between liver allograft recipients and controls (Table 1). Transplant patients were less likely to have private insurance, were more often white, and had more comorbidities, including hypertension. Only 50% were admitted to LT centers.
Table 1. Demographic and Admission Characteristics of Obstetric Hospitalizations According to Prior Liver Transplantation Status
|Demographics|| || || |
| Age (years)*||24 (22–31)||25 (22–31)||0.73|
|Health insurance|| || || |
| Medicare||13||1|| |
| Medicaid||44||44|| |
| Other/unknown||5||7|| |
|Race|| || || |
| African American||9||19|| |
| Hispanic||16||12|| |
| Other/unknown||28||31|| |
|Comorbid conditions|| || || |
| 1||59||14|| |
| ≥2||38||8|| |
|Hospital characteristics*|| || || |
| Liver transplant center||50||49||0.89|
| Urban, nonteaching||17||18|| |
| Urban, teaching||79||79|| |
| Region|| || || |
| Midwest||28||28|| |
| South||35||35|| |
| West||21||21|| |
|Admission characteristics|| || || |
| Elective admission||32||33||0.76|
| Year*|| || || |
| 2000–2005||60||60|| |
In the control group, 11% of obstetric hospitalizations were antepartum and for pregnancy complications not associated with delivery versus 29% in LT patients (P < 0.0001). The most frequent indications for antepartum admission in LT recipients (n = 60) were unspecified antenatal complications (23%), threatened preterm labor (12%), antenatal hepatic complications (12%), anemia (5%), and rejection or failure of the allograft (5%). In controls, the most common reasons for antepartum admission (n = 430) were threatened preterm labor (17%), unspecified antenatal complications (16%), genitourinary infections (8%), and diabetes (6%). After adjustments for confounding factors, transplanted women had a higher risk of antepartum admission [odds ratio (OR) = 2.27; 95% confidence interval (CI) = 1.59–3.25]. Other predictors of antepartum admission included nonwhite race (OR = 1.24; 95% CI = 1.01–1.53), nonprivate health insurance (OR = 1.55; 95% CI = 1.26–1.92), and more comorbidities (≥2 versus none: OR = 2.10; 95% CI = 1.58–2.80), including diabetes (OR = 5.08; 95% CI = 3.16–8.19).
Modes of Delivery
Unassisted (0.7% versus 9%; P < 0.0007) and assisted vaginal deliveries (eg, with forceps or vacuum extraction; 55% versus 69%; P = 0.0004) were less common, whereas the rate of cesarean delivery was higher (38% versus 24%; P = 0.0001) among LT patients versus controls (Table 2). However, after multivariate adjustment, transplantation was not associated with cesarean delivery (OR = 0.87; 95% CI = 0.60–1.27). Regardless of transplant status, independent predictors of cesarean section included older age (OR = 1.03; 95% CI = 1.02–1.05) and more comorbidities (≥2 versus none: OR = 2.40; 95% CI = 1.84–3.12), including gestational diabetes (OR = 1.47; 95% CI = 1.02–2.14) and hypertension (OR = 1.63; 95% CI = 1.29–2.07) and preexisting diabetes (OR = 2.65; 95% CI = 1.46–4.81). Patients hospitalized electively (OR = 1.21; 95% CI = 1.03–1.44), in the Northeast United States (versus the Western United States: OR = 1.29; 95% CI = 1.00–1.67), and during the latter years of the study (2000–2005 versus 1993–1999: OR = 1.33; 95% CI = 1.13–1.57) were more likely to undergo cesarean section. In LT recipients, cesarean deliveries increased from 27% between 1993 and 1999 to 46% between 2000 and 2005 (OR = 2.39; 95% CI = 1.18–4.83). In controls, the growth in cesarean delivery rates was smaller (20% versus 26%: OR = 1.35; 95% CI = 1.15–1.58).
Table 2. Unadjusted Prevalence of Pregnancy Outcomes According to Prior Liver Transplantation Status
|Obstetric outcomes|| || || || |
| Antepartum admission†||29.1||11.3||<0.0001||3.22 (2.35–4.41)|
| Cesarean delivery*||37.7||23.6||0.0001||1.95 (1.39–2.75)|
| Assisted delivery*||55.5||69.3||0.0004||0.55 (0.40–0.77)|
| Normal delivery*||0.7||8.6||0.0007||0.07 (0.01–0.53)|
| Multiple gestations*||1.4||2.6||0.59||0.53 (0.13–2.17)|
| Premature rupture of membranes*||5.5||4.5||0.60||1.22 (0.59–2.52)|
| Placenta previa*||1.4||0.5||0.18||2.76 (0.64–12.03)|
| Placental abruption*||2.7||1.8||0.34||1.56 (0.56–4.33)|
|Fetal complications|| || || || |
| Death†||6.3||2.0||0.0006||3.23 (1.77–5.89)|
| Spontaneous/missed abortions†||4.9||1.3||0.0009||3.79 (1.90–7.55)|
| Preterm*||27.4||10.8||<0.0001||3.13 (2.15–4.58)|
| Intrauterine growth restriction*||4.8||2.2||0.05||2.25 (1.02–4.95)|
| Fetal distress*||10.3||5.0||0.005||2.18 (1.25–3.79)|
| Congenital anomaly*||1.4||0.6||0.21||2.49 (0.58–10.74)|
| Any fetal complication*||39.7||17.8||<0.0001||3.04 (2.16–4.28)|
|Maternal complications|| || || || |
| Antepartum hemorrhage†||1.5||1.1||0.51||1.29 (0.40–4.18)|
| Postpartum hemorrhage*||7.5||3.2||0.009||2.47 (1.30–4.69)|
| Blood transfusion†||4.4||0.5||<0.0001||8.79 (3.97–19.44)|
| Peripartum infection†||1.9||1.8||0.79||1.10 (0.40–3.03)|
| Hypertension during pregnancy†||30.1||9.2||<0.0001||4.23 (3.09–5.81)|
| Preeclampsia||16.5||4.3||<0.0001||4.36 (2.93–6.49)|
| Gestational diabetes†||2.4||3.6||0.38||0.67 (0.27–1.65)|
| Venous thromboembolism†||0.5||0.2||0.29||3.30 (0.40–27.51)|
| Any maternal complication†‡||43.2||21.6||<0.0001||2.77 (2.08–3.68)|
Maternal and Fetal Pregnancy Outcomes
Table 2 includes the unadjusted prevalence of obstetric outcomes according to transplant status. Maternal deaths did not differ between cases and controls; 1 mother, a control patient, died (0.02%). Fetal deaths were more common in LT recipients than controls (6.3% versus 2.0%; P = 0.0006). The majority of these deaths were coded as spontaneous or missed abortions (77% of fetal deaths in LT recipients versus 65% in controls; P = 0.53). At least 1 fetal complication occurred more often among LT recipients (40% versus 18%; P < 0.0001). Specific complications that were more common among infants of transplant patients included prematurity (27% versus 11%; P < 0.0001), distress (10% versus 5%; P = 0.0005), and growth restriction (4.8% versus 2.2%; P = 0.05). Congenital anomalies (1.4% versus 0.6%; P = 0.25) and multiple gestations (1.4% versus 2.6%; P = 0.59) were similar between groups. In a multivariate analysis, transplantation was an independent risk factor for fetal complications (adjusted OR = 2.49; 95% CI = 1.68–3.70). Additional risk factors included cesarean delivery (OR = 1.74; 95% CI = 1.43–2.11), nonelective admission (OR = 1.49; 95% CI = 1.22–1.82), transfers from another institution (OR = 13.5; 95% CI = 7.75–23.5), and delivery during the earlier study period (1993–1999 versus 2000–2005: OR = 1.64; 95% CI = 1.38–1.96). Admission to an LT center (OR = 1.11; 95% CI = 0.92–1.33) and the number of comorbidities including diabetes (pregnancy-related and unrelated) and gestational hypertension were not significant (data not shown).
Maternal complications also occurred more frequently among LT recipients (43% versus 22%; P < 0.0001; Table 2). Specific complications that were more common among LT patients included postpartum hemorrhage (7.5% versus 3.2%; P = 0.009), blood transfusion (4.4% versus 0.5%; P < 0.0001), and gestational hypertension (30% versus 9%; P < 0.0001), including preeclampsia (17% versus 4%; P < 0.0001). Premature rupture of membranes, placental abruption and previa, antepartum bleeding, infections, gestational diabetes, and venous thromboembolism did not differ between groups. No patients developed ascites, encephalopathy, or variceal hemorrhage. After adjustments for confounding factors, maternal complications were more common among LT recipients than controls (OR = 2.63; 95% CI = 1.82–3.80). Additional independent predictors of maternal complications included cesarean delivery (OR = 1.63; 95% CI = 1.37–1.94), transfers from another institution (OR = 5.02; 95% CI = 3.07–8.20), and the number of comorbidities (≥2 versus none: OR = 1.56; 95% CI = 1.19–2.06), including preexisting diabetes (OR = 3.07; 95% CI = 1.71–5.53). Admission to an LT center or during the latter years of the study was not associated with maternal complications (data not shown).
LOS and Hospital Charges
Median LOS (interquartile range) was substantially longer in LT patients than controls for both vaginal deliveries [3 (2–5) versus 2 (1–3) days; P < 0.0001] and cesarean deliveries [4 (3–7) versus 4 (3–4) days; P < 0.0001]. Similarly, patients with a previous transplant accrued greater charges [median of $10,991 (interquartile range = $6142–20,616) versus $6853 (interquartile range = $4677–10,781); P < 0.0001]. After controlling for confounding factors, we found that adjusted LOS and hospital charges were 10% (95% CI = 1%–20%) and 23% (95% CI = 12%–35%) greater, respectively, in LT patients. Admission to an LT center was associated with greater adjusted LOS (11%; 95% CI = 7%–15%) but similar hospital charges (P = 0.79).
This is the first population-based study to describe obstetric outcomes among LT recipients. Our data demonstrate that maternal deaths and most adverse pregnancy outcomes do not differ significantly between transplant patients and controls. However, fetal deaths, antepartum admissions, and maternal and fetal complications overall are 2- to 3-fold greater and LOS and hospital charges are 10% to 23% higher in liver allograft recipients. Admission to an LT center, which surprisingly occurred in only 50% of patients, was not associated with maternal or fetal complications. Our findings add to the existing literature describing pregnancy outcomes post-LT (see Table 3) and provide useful data for counseling LT recipients who become pregnant or are contemplating pregnancy. The major strengths of our study are its population-based nature and large sample size. Although other large single-center series have been reported,3, 12 they are subject to referral bias (eg, studies from transplant centers). Similarly, pregnancy registries may be limited by a tendency to preferentially report adverse outcomes and/or underreport pregnancies overall. For example, the 2006 NTPR reported 151 deliveries between 1991 and 2006.9 By way of comparison, our data correspond to an estimated 736 deliveries in the United States between 1993 and 2005 after we account for the stratified sampling design of the NIS database.24 Another strength of our study is the use of a matched control group based on age, hospital, and year of admission, the latter to account for center-specific and temporal variations in outcomes. We also adjusted for differences in potentially confounding factors (eg, sociodemographics and comorbidities) between cases and controls. Because LT patients tend to have more comorbidities than the general childbearing population, it is vital to adjust for these differences when we are comparing obstetric outcomes.
Table 3. Obstetric Outcomes Among Liver Transplant Recipients from Selected Publications
|Registries|| || || || || || |
| NTPR9 (2006)||205 (151)||19||35||35||7||34|
| UK Transplant Registry10 (2007)||18 (11)||11||62||50||—||—|
|Case series|| || || || || || |
| Christopher et al.3 (2006)||70 (50)||19||40||27||17||20|
| Jain et al.12 (2003)||49 (49)||—||47||49||—||11|
| Nagy et al.14 (2003)||38 (24)||11||46||29||17||21|
| Patapis et al.13 (1997)||29 (15)||17||20||—||13||33|
| Raakow et al.16 (2001)||28 (21)||—||48||19||0||43|
| Wu et al.15 (1998)||22 (22)||—||32||14||5||14|
| Scantlebury et al.11 (1990)||19 (19)||—||63||53||5||32|
|Population-based studies|| || || || || || |
| Coffin et al. (2009)||206 (146)||5||38||27||5||30|
We observed a 2-fold risk of antepartum admission in liver allograft recipients, which supports an increased incidence of adverse pregnancy outcomes in this complex patient population. It is noteworthy that 17% of admissions were due to hepatic complications, including allograft failure and/or rejection (5%). In the 2006 NTPR, 7% of pregnancies were complicated by acute rejection, and 8% lost their grafts within 2 years of delivery; this was not significantly different from the nonpregnant transplant population.9 Other studies that have employed variable diagnostic criteria and inconsistently used biopsy for confirmation have reported rejection rates during pregnancy of 0% to 17% (Table 3).3, 11, 13–16 Regardless of transplant status, an increased risk of antepartum admission was observed among nonwhites and those without private health insurance, perhaps partly because of suboptimal outpatient care, as these factors have been associated with reduced access to prenatal care in the United States.25
LT patients also had a 2-fold risk of maternal complications, largely because of postpartum hemorrhage, a need for blood transfusion, and gestational hypertension. These hematological findings were independent of cesarean delivery (data not shown) and may be related to immunosuppression-related thrombocytopenia and/or anemia26 exacerbated by the hemodilution of pregnancy. As anemia was one of the most common indications for antepartum admission (5%), it is clearly an important issue that warrants close monitoring in pregnant LT patients.11, 14 In keeping with other reports,3, 9, 11, 12, 14 pregnancy-related hypertension (30% versus 9%), including preeclampsia (17% versus 4%), was more common among transplant recipients, perhaps because of the vasoconstrictive effects of immunosuppressants (ie, calcineurin inhibitors).27, 28 Although these drugs are also associated with diabetes in the nonpregnant population,29 rates of gestational diabetes were similar between groups (2%–4%).
Infants of LT patients also had a 3-fold risk of complications, most notably fetal death (6% versus 2% in controls). The majority of these deaths were coded as spontaneous or missed abortions, which are generally defined as fetal deaths occurring before the 22nd week of gestation.30 This rate of fetal loss among LT recipients is lower than previously reported (Table 3).3, 9, 10, 13, 14 For example, among 205 pregnancies in LT patients described in the 2006 NTPR, 19% and 2% ended in spontaneous abortions and stillbirths, respectively.9 At King's College Hospital, Christopher et al.3 reported a spontaneous abortion rate of 19% among 70 pregnancies in LT recipients. These differences are likely related to our reliance on ICD-9-CM coding for case ascertainment and the restriction of our analysis to women who were hospitalized, who represent a minority of cases.30 The most important additional adverse fetal outcomes in our study included preterm labor (27% versus 11%), fetal distress (10% versus 5%), and growth restriction (5% versus 2%). These findings are in keeping with previous reports.3, 9, 11, 12, 14 Because of limitations in the available data, we cannot discern the relative impact of specific maternal factors and immunosuppressant medications on the risk of these complications. However, it is important that the incidence of congenital anomalies—presumably overt malformations detectable in the early postpartum period—was similar between groups (∼1%). Although these data support the overall safety of immunosuppressants during pregnancy, we cannot rule out the underestimation of events due to the insensitivity of ICD-9-CM coding, a lack of rigorous surveillance by attending physicians (eg, in comparison with a prospective surveillance program), and the limitation of follow-up to the index hospitalization. Moreover, our sample size (n = 148 deliveries) is likely too small to detect uncommon events, such as congenital malformations described with in utero exposure to mycophenolate mofetil.31 Finally, other obstetrical complications such as infections, placental problems, and thromboembolic disorders did not differ between groups.
As reported by others (Table 3),3, 9, 11, 12, 14 we observed a higher rate of cesarean delivery among liver allograft recipients than controls (38% versus 24%). Because this difference was not significant after multivariate adjustment, it is likely that the higher rate of cesarean delivery was due to comorbidities and pregnancy complications (eg, preeclampsia and fetal distress) rather than LT itself.6 A previously unreported finding is the increasing frequency of cesarean delivery among transplant patients over time (from 27% during 1993–1999 to 46% during 2000–2005). This increase was greater than that observed in the nontransplant population and may reflect temporal changes in illness severity and/or patient and physician attitudes toward modes of delivery.
Our study has several limitations. First, as in all administrative database studies, the validity of ICD-9-CM coding must be considered. Although the pregnancy-related codes have not been validated in the NIS, they have been used in previous studies.32, 33 Second, we cannot exclude bias in the reporting of obstetric complications and specifically the propensity to preferentially record adverse events in transplant recipients because of their tendency to be sicker. Third, although some data regarding neonatal outcomes are routinely recorded in the mother's discharge abstract, the inability to formally link these discharge records may have limited our ascertainment of neonatal complications. Forth, the NIS database includes data on pregnancy-related hospitalizations only. Inferences regarding rates of conception and early pregnancy losses, which are probably more common among transplant recipients,3, 9, 14 cannot be made. In addition, the NIS does not include data on immunosuppression, the LT indication, or the interval between transplantation and conception, all of which may be associated with the risk of adverse events. For example, several studies have shown lower rates of complications, including spontaneous abortion, prematurity, and allograft rejection, in conceptions postponed at least 1 year after transplantation.3, 14 Finally, the NIS includes hospital outcomes only; important longer term outcomes such as childhood development are unavailable.
In summary, pregnancy in liver allograft recipients is not an uncommon event. Given the rapid return of fertility following transplantation, accurate family planning advice is essential. Our population-based study demonstrates that although the majority of pregnancies in LT patients have favorable outcomes, both mothers and their infants have an increased risk of complications. These data are consistent with other studies, including case series and voluntary registries, supporting the validity of our findings. Because mechanisms to reduce obstetric complications in LT recipients require evaluation, we encourage additional center-specific reports and continued enrollment in the NTPR. While we await further data, our findings support careful multidisciplinary management of pregnant LT recipients, including transplant hepatologists, maternal-fetal medicine specialists, and neonatologists when necessary.