Objective
Attention for recurrent preterm delivery has primarily focused on spontaneous subtypes with less known about indicated preterm delivery.
Study Design
In a retrospective cohort of consecutive pregnancies among 51,086 women in Utah (2002-2010), binary relative risk regression was performed to examine the risk of preterm delivery (PTD; <37 weeks) in the second observed delivery by PTD in the first, adjusting for maternal age, race/ethnicity, prepregnancy body mass index, insurance, smoking, alcohol and/or drug use, and chronic disease. Analyses were also performed stratified by prior preterm delivery subtype: spontaneous, indicated, or no recorded indication.
Results
There were 3836 women who delivered preterm in the first observed pregnancy (7.6%), of which 1160 repeated in the second (30.7%). Rate of recurrent PTD was 31.6% for prior spontaneous, 23.0% for prior indicated delivery, and 27.4% for prior elective delivery. Prior spontaneous PTD was associated with a relative risk (RR) of 5.64 (95% confidence interval [CI], 5.27–6.05) of subsequent spontaneous and RR of 1.61 (95% CI, 0.98–2.67) of subsequent indicated PTD. Prior indicated PTD was associated with an RR of 9.10 (95% CI, 4.68–17.71) of subsequent indicated and RR of 2.70 (95% CI, 2.00–3.65) of subsequent spontaneous PTD.
Conclusion
Prior indicated PTD was strongly associated with subsequent indicated PTD and with increased risk for subsequent spontaneous PTD. Spontaneous PTD had the highest rate of recurrence. Some common pathways for different etiologies of preterm delivery are likely, and indicated PTD merits additional attention for recurrence risk.
See related editorial, page 97
Pregnancy complications and adverse pregnancy outcomes often recur in subsequent pregnancies. Preterm delivery before 37 weeks of gestation is one common adverse outcome that repeats, in which women with a history have a 22% risk of preterm birth in a following pregnancy compared with a 9% risk in women without a history of prior preterm delivery. Yet even though women with a history of preterm birth have a 2.5-fold increase in spontaneous preterm delivery in the next pregnancy, the vast majority of women will still deliver at term. The tendency to recur increases with the number of prior preterm deliveries, earlier gestational age at prior delivery, and the order, with higher risk of subsequent preterm delivery if the immediately preceding birth was preterm. Yet our understanding of recurrence risk has primarily focused on spontaneous preterm deliveries or has not taken subtype into consideration.
Indicated preterm birth has been found to confer an increased risk for preterm delivery in subsequent pregnancies in one but not all studies. However, 3 of these studies relied on birth certificate data or birth registries, which are subject to misclassification of preterm birth subtype (eg, spontaneous vs indicated) and lacked detailed information on important risk factors. One study from a single institution investigated recurrent preterm birth less than 35 weeks of gestation.
These findings might not be relevant for all preterm births because most occur between 34 and 37 weeks. In addition, although gestational age at birth is known to be inversely related to risk of subsequent preterm delivery, with earlier gestational ages associated with higher risk in the next pregnancy, it is unknown whether we can predict not only whether but when a complication would recur and whether prior preterm birth subtype modifies the relationship. To address these critical data gaps, the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) performed a large retrospective cohort study with consecutive pregnancies.
Materials and Methods
The NICHD Consecutive Pregnancies Study collected detailed data on 114,679 pregnancies from 51,086 women delivering at 20 or more weeks of gestation from 20 hospitals in the state of Utah from 2002 to 2010. Internal review board approval was obtained at all participating institutions. All women contributed at least 2 deliveries, and births were consecutive across pregnancies based on increases in parity. However, 7712 women (15.1%) had a pregnancy loss prior to 20 weeks of gestation indicated by an increase in gravidity more than parity between pregnancies. Hospitals extracted detailed information from both the antepartum and labor and delivery summary electronic medical records.
Patient demographics, past medical history, reproductive and prenatal history, pregnancy, labor and delivery outcomes, and postpartum and neonatal information were mapped to predefined categories at the data coordinating center. The type of information available was as would be expected in typical clinical practice, although information on who (eg, nurse vs physician) entered data into the patient chart was not collected. For example, prenatal record included past medical history (eg, chronic hypertension, pregestational diabetes) and pregnancy complications (eg, gestational hypertension, preeclampsia, eclampsia, premature rupture of the membranes, intrauterine fetal growth restriction, etc).
Labor and delivery records included date and time of admission, cervical examination on admission, repeated cervical examinations during labor, labor and delivery characteristics (oxytocin, fetal presentation, etc), indications for induction (eg, fetal indication, maternal hypertensive disorder, premature rupture of membranes, postdate, etc), and indications for cesarean (eg, fetal malpresentation, prior cesarean, nonreassuring fetal heart tones, etc). Newborn records were linked to the neonatal intensive care unit records. International Classification of Diseases , ninth revision (ICD9) codes from maternal and newborn discharge summaries were linked to each delivery.
Potential clinical predictors of preterm birth were identified as summarized in the Institute of Medicine of the National Academies 2007 report from the Committee on Understanding Premature Birth and Assuring Healthy Outcomes. Three models with different sets of predictors were performed with the first including maternal characteristics of age; race/ethnicity; prepregnancy body mass index (BMI); insurance; smoking, alcohol, or drug use during pregnancy as recorded in the prenatal record (yes/no); and chronic diseases including diabetes, chronic hypertension, heart disease, renal disease, depression, seizure disorder, thyroid disease, and asthma. Chronic medical conditions were as recorded in the medical record and supplemented with discharge summary data using ICD9 codes.
The second model included the previously mentioned maternal characteristics as well as prior reproductive history including history of pregnancy loss (including miscarriage and terminations) calculated as gravidity minus parity greater than 1, history of stillbirth calculated as parity minus history of live birth, interpregnancy interval based on the number of days between the delivery and the last menstrual period of a subsequent pregnancy, and history of small for gestational age (birthweight less than the 10th percentile).
A final model included all of the previously mentioned maternal characteristics and prior reproductive history as well as pregnancy complications including genitourinary bacterial infection during pregnancy (sexual transmitted disease or urinary tract infection), vaginal bleeding, placenta previa or accreta, and uterine anomaly as identified by ICD-9 codes (752.2, doubling of uterus [didelphic uterus] or 752.3, other anomalies of uterus [bicornuate, unicornis, uterus with only 1 functioning horn]).
Categories of preterm birth subtypes were created using the following algorithm previously published by our group : induction or prelabor cesarean delivery recorded in the medical record was used to identify the nonspontaneous precursors for delivery. A woman was considered to have presented in spontaneous labor if she did not have an induction or prelabor cesarean delivery. Women with spontaneous labor and other pregnancy complications (eg, preeclampsia) were included only in the spontaneous preterm delivery category. Women with premature rupture of the membranes and not in labor were included as preterm premature rupture of membranes (PPROM). If a women presented with both PPROM and in spontaneous labor, she was counted only once in the spontaneous labor category.
The results for analyses of spontaneous labor and PPROM were similar, so we elected to combine these categories because of the small numbers in certain analyses. If a woman did not present in spontaneous labor or with PPROM, we then identified all potential maternal, fetal, or obstetrical complications of pregnancy and included these in the indicated category. The rationale for including all complications was if a woman had an induction or prelabor cesarean without an indication listed and the pregnancy had a complication (eg, preeclampsia), we wanted to be conservative and assume that the most likely reason for delivery was preeclampsia rather than classify as a medically unnecessary delivery. The final category included labor inductions or cesarean deliveries recorded as elective by the site with no other obstetrical, fetal, or maternal conditions as well those deliveries with no recorded indication.
The analysis was restricted to singleton pregnancies. Maternal characteristics upon entry to the cohort and pregnancy characteristics by delivery were summarized. The remaining analyses were limited to the first 2 or 3 pregnancies in the data set. A scatterplot was created of gestational age for the first delivery vs gestational age for the second delivery with a Loess smooth curve overlaid. Binary relative risk regression (binary regression with a log-link function) was performed to calculate relative risk (RR) and 95% confidence intervals (CIs) of preterm birth in the second observed delivery for the category of gestational age for the first observed delivery. Because all women in the data set had at least 2 pregnancies by design, risk for women with more or fewer pregnancies could have been different from the extent that their measured and unmeasured characteristics varied. Therefore, a sensitivity analysis was performed limiting the analysis to women who contributed at least 3 pregnancies to the dataset. Relative risk regression was performed to examine the risk of preterm birth in the third delivery by preterm birth in the first and second deliveries.
We also fit models with an interaction term to examine the interaction of preterm birth in the first and second delivery on the occurrence of preterm in the third delivery. The models with interaction terms were adjusted for the same covariates as mentioned in previous text for the additive model. Cumulative incidence curves were estimated by prior delivery timing (20 to <24, 24 to <28, 28 to <34, 34 to <37, and ≥37 weeks of gestation) using Kaplan-Meier estimation. The incidence curves were presented for the gestational age in the second delivery by categories of gestational age in the first delivery.
Results
There were 51,066 women with a total of 114,639 singleton pregnancies. Maternal characteristics upon entry to the cohort study and pregnancy characteristics by delivery are presented in Table 1 . The majority of women (39,954, 78.2%) contributed only 2 pregnancies, an additional 9792 (19.2%) contributed 3 pregnancies, and 1320 women contributed 4 or more pregnancies (2.6%). There were 3836 women who delivered preterm in the first pregnancy (7.6%), of which 1160 (30.7%) repeated in the second. The timing of delivery in the first delivery was correlated with timing of delivery in the subsequent pregnancy; however, there was substantial variation ( Figure 1 ).
| Characteristic | Number of women (%) a (n = 51,066) |
|---|---|
| Age, mean (SD), y | 25.6 (4.5) |
| Younger than 18 | 1093 (2.1) |
| 18-34 | 48,099 (94.2) |
| 35 or older | 1874 (3.7) |
| BMI, mean (SD) b | 24.3 (5.3) |
| Race/ethnicity | |
| Non-Hispanic white | 44,054 (86.3) |
| Non-Hispanic black | 223 (0.4) |
| Hispanic | 5431 (10.6) |
| Asian/Pacific Islander | 1088 (2.1) |
| Other/unknown | 270 (0.5) |
| Marital status | |
| Married | 44,032 (86.2) |
| Divorced/widowed | 692 (1.4) |
| Single | 6333 (12.4) |
| Unknown | 9 (0.02) |
| Insurance status | |
| Private | 37,370 (73.8) |
| Public | 13,237 (26.2) |
| Gravidity, median (range) | 2 (1–24) |
| Parity, median (range) | 0 (0–14) |
| 0 | 27,730 (59.5) |
| 1 | 11,704 (25.1) |
| ≥2 | 7147 (15.3) |
| Smoker | 1253 (2.5) |
| Alcohol use | 871 (1.7) |
| Illicit drug use | 103 (0.2) |
| Chronic medical disease | 8427 (16.5) |
| Pregestational diabetes | 606 (1.2) |
| Chronic hypertension | 277 (0.5) |
| Uterine anomaly | 116 (0.2) |
| Genitourinary bacterial infection | |
| Sexually transmitted disease | 540 (1.1) |
| Urinary tract infection | 1369 (2.7) |
| Vaginal bleeding | 504 (1.0) |
| Interpregnancy interval, median (range), d | 562 (15–2410) |
| Prior adverse pregnancy outcome c | |
| Miscarriage or termination | 10,817 (21.2) |
| Preterm birth | 2019 (4.0) |
| Number of pregnancies | |
| 2 | 39,954 (78.2) |
| 3 | 9792 (19.2) |
| 4 | 1247 (2.4) |
| 5 | 71 (0.1) |
| 6 | 2 (0.0) |
b Calculated as weight in kilograms divided by height in meters squared
c Prior outcome only for deliveries in which gravidity is greater than 1. Miscarriage or termination was calculated as yes if gravidity-parity is greater than 1 (could include neonatal deaths, stillbirths, abortions, and miscarriages).
The earlier the gestational age of the first delivery, the higher the risk of preterm birth of less than 37 weeks of gestation in the subsequent delivery, in which the trend for gestational age was highly significant ( P < .0001) ( Table 2 ). However, although the highest risk for subsequent preterm birth less than 37 weeks was with a prior preterm birth at 24-27 weeks of gestation (RR, 7.03; 95% CI, 5.77–8.57), the RR were overall high for all prior preterm birth gestational age categories and the confidence intervals overlapped.
| Gestational age at first delivery, wks | Total n (%) | Preterm in second birth, n (%) | Preterm birth <37 weeks in second delivery | |
|---|---|---|---|---|
| Unadjusted RR (95% CI) | Adjusted RR a (95% CI) | |||
| ≥37 | 46771 (92.4) | 2630 (5.7) | Referent | Referent |
| 34 to <37 | 2950 (5.8) | 838 (28.9) | 5.07 (4.73–5.42) | 4.81 (4.48–5.15) |
| 28 to <34 | 607 (1.2) | 226 (37.9) | 6.63 (5.95–7.40) | 5.98 (5.37–6.66) |
| 24 to <28 | 152 (0.3) | 61 (40.1) | 7.03 (5.77–8.57) | 6.42 (5.33–7.74) |
| 20 to <24 | 127 (0.3) | 35 (27.8) | 4.87 (3.66–6.47) | 4.88 (3.66–6.50) |
a Models were adjusted for maternal age, race/ethnicity, prepregnancy body mass index, insurance, smoker, alcohol, illicit drug use, and chronic medical disease.
These findings were supported by the second delivery cumulative incidence curves, which were ordered by prior delivery gestational age category until around 28 weeks, after which the degree of preterm was not as influential as any history of preterm birth ( Figure 2 , A). However, although the degree of preterm birth was highly associated with future risk, history still was not very predictive of either subsequent preterm birth or subsequent preterm gestational age at delivery. For example, if a woman had a prior preterm delivery at 28 weeks, the cumulative incidence of any preterm delivery less than 37 weeks was 37.9%, but the incidence of delivery 28 weeks or less was only 10.4%. Thus, she still had significant risk of preterm delivery throughout the next pregnancy, and there was no gestational age cutoff at which the risk was no longer increased.
Because the number of pregnancies per woman in the study could have resulted in differences in their measured and unmeasured characteristics, a sensitivity analysis limited to women who contributed at least 3 pregnancies to the data set was performed. The results of the analyses on this substantially smaller subcohort along with the analyses on the first 2 pregnancies on the full cohort suggest that the underlying recurrence risk was similar, regardless of the number of pregnancies that a woman contributed.
After adjusting for clinical characteristics at baseline in pregnancy, the RR for subsequent preterm birth were slightly attenuated but overall the same ( Table 2 ). Furthermore, models that included these maternal characteristics as well as prior obstetrical history including interpregnancy interval, history of miscarriage or termination, and history of small for gestational age less than the 10th percentile as well as pregnancy complications including genitourinary bacterial infection during pregnancy, vaginal bleeding, placenta previa or accreta, and uterine anomaly had similar results (data not shown). Therefore, most of the risk of preterm delivery was captured by the unadjusted model signifying that prior history of preterm birth was the single most important predictor of subsequent preterm birth less than 37 weeks of gestation. In spite of these high relative risks, 60-72% of the women still delivered at term in the subsequent pregnancy.
Three consecutive pregnancies
The risk of preterm birth less than 37 weeks of gestation in the third delivery was greater than 10-fold higher for women with 2 prior preterm deliveries compared with women with 2 prior term deliveries ( Table 3 ). For women with a history of 1 prior preterm and 1 prior term delivery, the risk for subsequent preterm birth in the third pregnancy was higher, 5.5-fold vs 3.5-fold, when the preterm birth occurred in the immediate preceding pregnancy compared with a preterm birth in the first pregnancy.
