Advertisement

Recurrence risk of preterm birth in subsequent singleton pregnancy after preterm twin delivery

      Objective

      The purpose of this study was to investigate the recurrence risk of preterm birth (<37 weeks' gestation) in a subsequent singleton pregnancy after a previous nulliparous preterm twin delivery.

      Study Design

      We included 1957 women who delivered a twin gestation and a subsequent singleton pregnancy from the Netherlands Perinatal Registry. We compared the outcome of subsequent singleton pregnancy of women with a history of preterm delivery to the pregnancy outcome of women with a history of term twin delivery.

      Results

      Preterm birth in the twin pregnancy occurred in 1075 women (55%) vs 882 women (45%) who delivered at term. The risk of subsequent spontaneous singleton preterm birth was significantly higher after preterm twin delivery (5.2% vs 0.8%; odds ratio, 6.9; 95% confidence interval, 3.1–15.2).

      Conclusion

      Women who deliver a twin pregnancy are at greater risk for delivering prematurely in a subsequent singleton pregnancy.

      Key words

      In many developed countries, twin birth rates have increased drastically during the past decades. In the United States, the twin rate climbed from 1.9% in 1980 to 3.2% in 2006.
      • Chauhan S.P.
      • Scardo J.A.
      • Hayes E.
      • Abuhamad A.Z.
      • Berghella V.
      Twins: prevalence, problems, and preterm births.
      Similar trends were found in other developed countries.
      • Imaizumi Y.
      Trends of twinning rates in ten countries, 1972-1996.
      The increase in twin birth rates is caused mainly by the increase in the use of assisted reproductive techniques and increasing maternal age.
      • Chauhan S.P.
      • Scardo J.A.
      • Hayes E.
      • Abuhamad A.Z.
      • Berghella V.
      Twins: prevalence, problems, and preterm births.
      • Eriksson A.W.
      • Fellman J.
      Temporal trends in the rates of multiple maternities in England and Wales.
      For Editors' Commentary, see Contents
      See related editorial, page 241
      Twin pregnancies are associated with higher risks of various pregnancy complications such as preeclampsia, intrauterine growth restriction, and preterm birth.
      • Martin J.A.
      • Hamilton B.E.
      • Ventura S.J.
      • et al.
      Births: final data for 2009.
      These preterm births either can be a result of intervention in the case of previously mentioned obstetrics complications or can occur spontaneously. Preterm birth, in its turn, is the most important risk factor for perinatal morbidity and death in developed countries.
      • Ananth C.V.
      • Vintzileos A.M.
      Epidemiology of preterm birth and its clinical subtypes.
      • Wen S.W.
      • Smith G.
      • Yang Q.
      • Walker M.
      Epidemiology of preterm birth and neonatal outcome.
      This is mostly due to respiratory immaturity, intracranial hemorrhages, and infections.
      • Slattery M.M.
      • Morrison J.J.
      Preterm delivery.
      Besides having a twin pregnancy, a history of previous preterm birth is the most important risk factor for spontaneous preterm birth. This recurrence risk was demonstrated particularly for singleton pregnancies in women with a preceding preterm singleton delivery.
      • Adams M.M.
      • Elam-Evans L.D.
      • Wilson H.G.
      • Gilbertz D.A.
      Rates of and factors associated with recurrence of preterm delivery.
      • Goldenberg R.L.
      • Iams J.D.
      • Mercer B.M.
      • et al.
      The preterm prediction study: the value of new vs standard risk factors in predicting early and all spontaneous preterm births: NICHD MFMU Network.
      Less is known about the recurrence risk of preterm birth after a preceding twin pregnancy. The few studies that have reported on this phenomenon had contradictory findings and had access to relatively small samples.
      • Bloom S.L.
      • Yost N.P.
      • McIntire D.D.
      • Leveno K.J.
      Recurrence of preterm birth in singleton and twin pregnancies.
      • Facco F.L.
      • Nash K.
      • Grobman W.A.
      Are women who have had a preterm twin delivery at greater risk of preterm birth in a subsequent singleton pregnancy?.
      • Menard M.K.
      • Newman R.B.
      • Keenan A.
      • Ebeling M.
      Prognostic significance of prior preterm twin delivery on subsequent singleton pregnancy.
      • Rydhstroem H.
      Gestational duration in the pregnancy after a preterm twin delivery.
      • Rafael T.J.
      • Hoffman M.K.
      • Leiby B.E.
      • Berghella V.
      Gestational age of previous twin preterm birth as a predictor for subsequent singleton preterm birth.
      Therefore, we aim to investigate the recurrence risk of a spontaneous preterm birth in subsequent singleton pregnancy after previous preterm twin delivery in a nationwide database.

      Materials and Methods

      Dataset

      This study was performed in a prospective nationwide cohort with the use of The Netherlands Perinatal Registry (PRN). The PRN consists of population-based data that contain information on pregnancies, deliveries, and readmissions until 28 days after birth. The PRN database is obtained by a validated linkage of 3 different registries: the midwifery registry, the obstetrics registry, and the neonatology registry of hospital admissions of newborn infants.
      • Meray N.
      • Reitsma J.B.
      • Ravelli A.C.
      • Bonsel G.J.
      Probabilistic record linkage is a valid and transparent tool to combine databases without a patient identification number.
      • Tromp M.
      • Ravelli A.C.
      • Meray N.
      • Reitsma J.B.
      • Bonsel G.J.
      An efficient validation method of probabilistic record linkage including readmissions and twins.
      The coverage of the PRN registry is approximately 96% of all deliveries in The Netherlands. It contains pregnancies of ≥22 weeks' gestation and a birthweight of ≥500 g and is used primarily for an annual assessment of the quality indicators of obstetric care.

      Longitudinal linkage

      The records that are included in the PRN registry are entered at the child's level. There is no unique maternal identifier available in the registry to follow-up on outcomes of subsequent pregnancies of the same mother. Therefore, we performed a longitudinal probabilistic linkage procedure in which we linked records of children of the same mother to create a mother identifier. We subjected all children from second deliveries (n = 509,559) who were registered in the PRN registry to linkage with their siblings who were born during a first delivery that had been registered in the PRN registry. The linkage was based on the variables of the birth date of mother, the birth date of a (previous) child, and the postal code of mother (Appendix). The final linked cohort with complete data on first and second deliveries of the same mother consisted of 272,551 women and 545,102 (2 × 272,551) deliveries.

      Inclusion and exclusion criteria

      From our linked cohort, we included all women who delivered a singleton pregnancy (second delivery) after a previous twin pregnancy (first delivery) in The Netherlands between January 1, 1999, and December 31, 2007. We excluded all cases with major congenital anomalies and all cases with antepartum fetal death. Preterm birth was defined as birth at <37 completed weeks of gestation. We excluded iatrogenic preterm births in the subsequent singleton pregnancies because we were interested only in the subsequent risk of spontaneous preterm birth.

      Statistics

      We compared women with a preterm twin delivery with women with a term twin delivery.
      For these 2 groups, we compared demographic and obstetric baseline characteristics like maternal age (mean ± SD), white maternal ethnicity (yes vs no), socioeconomic status (low (<p25) vs >p25), living in a deprived area (yes vs no), use of assisted reproductive techniques (yes vs no), and pregnancy interval (mean ± SD).
      We subdivided previous preterm deliveries into iatrogenic and spontaneous deliveries. Furthermore, we subdivided previous preterm birth into 3 subgroups (22+0-29+6 weeks' gestation, 30+0-33+6 weeks' gestation, and 34+0-36+6 weeks gestation). Univariate analyses were performed with the Student t test for normally distributed continuous variables and Fisher exact test for categoric variables. Normality of continuous variables was assessed by visual inspection of Q-Q plots. All statistical tests were 2-sided; a probability value of .05 was chosen as the threshold for statistical significance. We measured the association between history of preterm birth and subsequent risk of spontaneous preterm birth by calculating an adjusted odds ratio (aOR). We adjusted only for variables that appeared to be distributed unequally in the baseline characteristics of the study population. The probabilistic linkage procedure was performed with the R statistical software environment (version 2.13.1; R Foundation for Statistical Computing, Vienna, Austria), and the data were analyzed with the SAS statistical software package (version 9.2; SAS Institute Inc, Cary, NC).

      Results

      To determine which children had similar mothers, the PRN dataset was divided into 2 datasets. Dataset A contained records of second deliveries (n = 509.559), and dataset B contained records (n = 667.053) of first deliveries. By performing a probabilistic record linkage procedure, we determined which second delivery from dataset A belonged to a first delivery from dataset B. After the longitudinal linkage procedure (Appendix), we were able to identify 272,551 pairs of first and second deliveries. The linked dataset consisted of 254,776 singleton-singleton pairs (97.7%), 4071 singleton-twin pairs (1.6%), 57 twin-twin pairs (0.02%), and 2097 (0.8%) mothers who had a twin delivery that was followed by a singleton delivery.
      We selected the 2097 women with a twin delivery followed by a singleton delivery for our study. We excluded mothers with iatrogenic preterm births in the second pregnancy (1.8%), severe congenital anomalies in first or second pregnancy (1.8% and 1.1%, respectively), and antepartum fetal death (2.1% and 0.4%, respectively). Our final dataset consisted of 1957 women.
      Baseline characteristics of this cohort are presented in Table 1. In the twin pregnancy group, 1075 women (55%) delivered at <37 completed weeks of gestation. In most cases, these preterm births were a result of obstetric intervention (n = 597; 56%) but occurred spontaneously in the remaining 478 women (44%). Demographic characteristics of the women with preterm (n = 1075) and term (n = 882) twin deliveries were comparable when we considered maternal age, socioeconomic status, living in a deprived area, and the use of artificial reproductive technology. Nonetheless, there were significantly fewer women with a white maternal ethnicity in the group with preterm twin deliveries (88.7% vs 91.4%; P < .05). The time interval to the subsequent singleton pregnancy was statistically significantly shorter in the women who delivered their twins preterm (33 vs 36 months; P < .001). As expected, the mean gestational age was also significantly different between the 2 groups.
      TABLE 1Baseline maternal characteristics of the cohort (n = 1957) stratified by gestational age at the twin delivery
      Characteristics of the twin deliveryPreterm twin delivery at <37 weeks (n = 1075)Term twin delivery at ≥37 weeks (n = 882)P value
      Gestational age at twin delivery, wk
      Data are given as mean ± SD.
      32.5 ± 3.938.0 ± 1.1< .0001
      Maternal age at twin delivery, y
      Data are given as mean ± SD.
      29.1 ± 4.029.2 ± 4.1.54
      White maternal ethnicity, n (%)982 (88.7)782 (91.4)< .05
      Low socioeconomic status, n (%)207 (19.3)185 (21.0).36
      Living in a deprived area, n (%)44 (4.1)43 (4.9).44
      Artificial reproductive technology, n (%)521 (48.5)426 (48.3).96
      Interval to subsequent singleton delivery, mo
      Data are given as mean ± SD.
      33 ± 1736 ± 16< .001
      Schaaf. Recurrence risk of preterm singleton birth after twin delivery. Am J Obstet Gynecol 2012.
      a Data are given as mean ± SD.
      Of the 1075 women who had a preterm twin delivery, 56 women (5.2%) had a spontaneous preterm birth in the subsequent singleton pregnancy, and 1019 women (94.8%) women delivered at term. The spontaneous singleton preterm birth rate in the 882 women who delivered their twins at term was 0.8% (n = 7). Delivery of preterm twins was associated therefore with a significant increased risk of spontaneous preterm birth in a subsequent singleton pregnancy (aOR, 6.9; 95% CI, 3.1–15.2). Table 2 shows the subdivision in spontaneous vs iatrogenic preterm twin delivery and that the increased risk of subsequent singleton preterm birth is even higher after a spontaneous preterm twin delivery (aOR, 9.9; 95% CI, 4.4–22.4) instead of an iatrogenic preterm twin delivery. Nevertheless, even after an iatrogenic preterm twin delivery, there is still an increased risk of spontaneous preterm birth in the next singleton pregnancy (aOR, 4.6; 95% CI, 1.9–10.8).
      TABLE 2Risk of preterm birth in subsequent singleton pregnancy
      Twin delivery (n = 1957)Subsequent singleton delivery (n = 1957)
      Spontaneous preterm birthOdds ratio
      Adjusted for maternal ethnicity and pregnancy interval.
      (95% CI)
      Gestational age at deliverynN%
      Total term delivery at ≥37 weeks88270.8Reference
      Total preterm delivery at <37 weeks1075565.26.9 (3.1–15.2)
      Spontaneous preterm delivery, wk
       <37478357.39.9 (4.4–22.4)
       34+0–36+619973.54.6 (1.6–13.3)
       30+0–33+611697.810.7 (3.9–29.5)
       22+0–29+61631911.717.8 (7.2–44.1)
      Iatrogenic preterm delivery
       <37597213.54.6 (1.9–10.8)
       34+0–36+640792.22.8 (1.04–7.6)
       30+0–33+614474.96.4 (2.2–18.5)
       22+0–29+646510.916.2 (4.9–54.1)
      CI, confidence interval.
      Schaaf. Recurrence risk of preterm singleton birth after twin delivery. Am J Obstet Gynecol 2012.
      a Adjusted for maternal ethnicity and pregnancy interval.
      Table 2 shows that, for both spontaneous and iatrogenic preterm twin deliveries, the recurrence risk also depends on the gestational age at the time of preterm twin delivery. The risk of preterm birth increases as the gestational age at preterm twin delivery decreases. The ORs in Table 2 are adjusted for maternal ethnicity and pregnancy interval.

      Comment

      We investigated the risk of spontaneous preterm (<37 weeks' gestation) singleton birth in women with a history of twin delivery. We found that the risk of subsequent singleton preterm birth is increased significantly after a previous preterm twin delivery when compared with a previous term twin delivery. Twin gestation is thus not only a risk factor for preterm birth in the current pregnancy but also accounts for an increased risk (5.2% vs 0.8%) in a subsequent singleton pregnancy.
      Our study was based on data from a large, well-maintained population-based national perinatal registry. The vast majority of the caregivers contribute to the PRN registry; therefore, it comprises approximately 96% of all pregnancy and birth characteristics in The Netherlands. The 4% missing birth data are due to 1-2% nonreporting general practitioners and 2-3% nonreporting midwives. Because threatened preterm delivery and multiple gestation are an indication for referral to an obstetrician and the registration by obstetricians is nearly complete (>99%), we would not have missed many cases because of nonreporting. Because of the magnitude of the PRN database, we did not perform an a priori power calculation. Unfortunately, the PRN does not contain information on possible confounders such as chorionicity, tobacco use, highest level of education, a history of cervical surgery, cervical length measurements, and the use of progesterone or cerclage.
      For our analyses, we performed a probabilistic linkage method to follow up with mothers in a subsequent pregnancy. Of the 509,559 second deliveries in the PRN registry, we were able to find the matching first delivery in 272,551 (53%) cases. Nonlinkage could be due to missing values of the linkage variables and by the fact that the first child was born before the start of the PRN registry in 1999. Finally, because “postal code of mother” is one of the linkage variables, changes of home address over time will also have led to nonlinkage. We found that our linked dataset of 272,551 women were comparable with the national figures on the level of demographic characteristics (eg, maternal age) and pregnancy outcomes (eg, pregnancy length, preterm birth rates, and congenital malformations). Only the twin pregnancy rate and the perinatal mortality rate in the linked dataset seem to be different from the original dataset in the first pregnancy (Appendix). However, we do not think that the nonlinked twin pregnancies have influenced our results to a large degree because nonlinkage is not related to gestational age at nulliparous twin delivery or to our primary outcome measure.
      The 5 previous publications on this subject have reported conflicting results and conclusions. However, they are based on relatively small sample sizes of <300 women. The most recent published study was performed by Rafael et al,
      • Rafael T.J.
      • Hoffman M.K.
      • Leiby B.E.
      • Berghella V.
      Gestational age of previous twin preterm birth as a predictor for subsequent singleton preterm birth.
      who investigated 255 women in a retrospective study. The risk of spontaneous preterm singleton birth was 11.1% after previous spontaneous twin birth vs 1.8% after previous term twin birth (OR, 6.81; 95% CI, 1.53–30.29). However, they did not find a significantly increased risk when the preterm twins were born between 34+0-36+6 weeks' gestation. The latter is in contrast with our findings but probably is due to a lack of statistical power in the study by Rafael et al.
      • Rafael T.J.
      • Hoffman M.K.
      • Leiby B.E.
      • Berghella V.
      Gestational age of previous twin preterm birth as a predictor for subsequent singleton preterm birth.
      Our results are also consistent with those presented by Facco et al,
      • Facco F.L.
      • Nash K.
      • Grobman W.A.
      Are women who have had a preterm twin delivery at greater risk of preterm birth in a subsequent singleton pregnancy?.
      who investigated 167 women with twin deliveries followed by a singleton gestation in a retrospective cohort study. The OR for spontaneous preterm birth in the singleton pregnancy was 5.0 (95% CI, 1.1–22.9) for women with a preceding preterm twin delivery compared with women with a term twin delivery. Similar results were found by Menard et al,
      • Menard M.K.
      • Newman R.B.
      • Keenan A.
      • Ebeling M.
      Prognostic significance of prior preterm twin delivery on subsequent singleton pregnancy.
      whose retrospective cohort consisted of 144 women, with a relative risk of subsequent singleton preterm birth after preterm twin delivery of 2.87 (95% CI, 1.02–8.09). All 3 studies also included women who had delivered before their index twin pregnancy, whereas in our study the women were nulliparous at the time of twin pregnancy. Another conclusion, compared with ours, was drawn by Bloom et al,
      • Bloom S.L.
      • Yost N.P.
      • McIntire D.D.
      • Leveno K.J.
      Recurrence of preterm birth in singleton and twin pregnancies.
      who concluded that previous preterm delivery (<35 weeks' gestation) did not significantly increase the risk of subsequent singleton preterm birth (OR, 1.9; 95% CI, 0.46–8.14). The analyses were performed in a cohort of only 82 women, thus probably lacked sufficient power to detect the difference in the risk of subsequent preterm delivery. In another study on this topic, the authors also concluded that, in most cases, a singleton pregnancy will proceed to term, irrespective of the gestational length of the preceding twin pregnancy.
      • Rydhstroem H.
      Gestational duration in the pregnancy after a preterm twin delivery.
      However, the authors base their conclusion on a frequency table without statistically comparing the 2 groups. They also analyzed previous preterm births as a whole, instead of subdividing them into iatrogenic and spontaneous preterm births. As our results showed, the recurrence risk after iatrogenic preterm birth is much smaller than in the case of a spontaneous preterm twin delivery. Combining iatrogenic and spontaneous preterm birth might have led to their different conclusion.
      Previous studies showed that the overall incidence of preterm birth is 6.0% in singleton pregnancies and 48.1% in twin pregnancies.
      • Schaaf J.
      • Mol B.
      • Abu-Hanna A.
      • Ravelli A.
      Trends in preterm birth: singleton and multiple pregnancies in the Netherlands, 2000-2007.
      Furthermore, there is a relatively expectant approach towards obstetric interventions in the preterm period in The Netherlands,
      • Schaaf J.
      • Mol B.
      • Abu-Hanna A.
      • Ravelli A.
      Trends in preterm birth: singleton and multiple pregnancies in the Netherlands, 2000-2007.
      • Zeitlin J.
      • Di L.D.
      • Blondel B.
      • et al.
      Variability in caesarean section rates for very preterm births at 28-31 weeks of gestation in 10 European regions: results of the MOSAIC project.
      which is also reflected in the low incidence of iatrogenic preterm births in our study population (30.1% in the nulliparous twin delivery and 1.8% in the excluded multiparous singleton delivery). The expectant approach is embedded in Dutch guidelines. For instance, in the case of premature prelabor rupture of membranes, the decision often is made to provide expectant monitoring until 37 weeks gestation instead of routine induction of labor.
      • Van der Ham D.P.
      • Nijhuis J.G.
      • Mol B.W.
      • et al.
      Induction of labour versus expectant management in women with preterm prelabour rupture of membranes between 34 and 37 weeks (the PPROMEXIL-trial).
      The low incidence figures in The Netherlands thus are not a result of a healthier population of pregnant women (which would limit the external validity) but rather a result of different doctors' behavior.
      To the best of our knowledge, our study has the largest sample size in the investigation of recurrence risk of singleton preterm birth after preterm twin delivery. The evidence of an increased risk of singleton preterm birth after a previous preterm singleton delivery is substantial and consistent.
      • Adams M.M.
      • Elam-Evans L.D.
      • Wilson H.G.
      • Gilbertz D.A.
      Rates of and factors associated with recurrence of preterm delivery.
      • Goldenberg R.L.
      • Iams J.D.
      • Mercer B.M.
      • et al.
      The preterm prediction study: the value of new vs standard risk factors in predicting early and all spontaneous preterm births: NICHD MFMU Network.
      We have now demonstrated a similar increased recurrence risk after preterm twin delivery. This increased risk is found after spontaneous and iatrogenic preterm delivery. The latter might be explained by the presence of risk factors (eg, maternal age of maternal ethnicity) that contributed both to the need of medical intervention in the first twin pregnancy and to the pathogenesis of spontaneous preterm birth in the next singleton pregnancy. Given the very low risk (0.8%) of spontaneous preterm birth after previous term twin delivery, one could also state the inversed conclusion: the risk of subsequent singleton preterm birth is decreased significantly after a previous term twin delivery when compared with a previous preterm twin delivery.
      The increased risk of preterm birth in twin pregnancies, which are often a result of artificial reproductive technology, thus not only is applicable to the current pregnancy but also impacts subsequent singleton pregnancies. These results can help clinicians to counsel their patients with a history of spontaneous or iatrogenic preterm delivery of twins and quantify their recurrence risks for spontaneous preterm birth.
      Previous singleton preterm birth is often an indication for the use of 17-alpha-hydroxyprogesterone in the next singleton pregnancy as a preventive measure for the recurrence of preterm birth.
      • Meis P.J.
      • Klebanoff M.
      • Thom E.
      • et al.
      Prevention of recurrent preterm delivery by 17 alpha-hydroxyprogesterone caproate.
      With these and previous findings, one should investigate the effectiveness of 17-alpha-hydroxyprogesterone in singleton pregnancies after preterm twin deliveries.

      Acknowledgments

      We thank all Dutch midwives, obstetricians, neonatologists, and other perinatal health care providers for the registration of perinatal information AND the Foundation of The Netherlands Perinatal Registry (www.perinatreg.nl) for permission to use the registry data.

      Appendix

      To determine which children had similar mothers, the The Netherlands Perinatal Registry dataset was divided into 2 datasets. Dataset A contained records (n = 509,559) of second deliveries; dataset B (n = 667,053) contained records of first deliveries. By performing a probabilistic record linkage procedure,
      • Fellegi I.P.
      • Sunter A.B.
      A theory for record linkage.
      we determined which second delivery from dataset A belonged to a first delivery from dataset B. Similarly to Tromp et al,
      • Tromp M.
      • Reitsma J.B.
      • Ravelli A.C.J.
      • Méray N.
      • Bonsel G.J.
      Record linkage: making the most out of errors in linking variables.
      records from A and B that belong to the same mother are called matches and nonmatches otherwise. Note that from the information on the deliveries that belonged to each other, we determined which children had similar mothers.
      We defined 3 linking variables (v1, v2, and v3) that were compared with each other to determine whether the ith delivery from dataset A belonged to the same mother as the jth delivery from dataset B. First, postal code (v1) and the date of birth (v2) of the mother were registered in both datasets. In addition to these characteristics of the mother, the date of the previous (first) delivery was registered in dataset A. This variable was compared with the date of the delivery that was registered in dataset B (v3). Note that the outcome of the comparison of the pth linking variable could be either an agreement or disagreement (ie, γijp = 1 or γijp = 0) or missing (ie, δijp = 1 or δijp = 0).
      To calculate the posterior probability, the following likelihood formula was maximized:
      ij=1nm{πp=13[μpγijp(1μp)(1γijp)]δijp+(1π)p=13[υpγijp(1υp)(1γijp)]δijp}


      where μp and υp are the probabilities of agreement of the pth variable among respective matches and nonmatches. In addition, π is the relative frequency of matches among the nm records-pairs.
      The estimated parameters that were derived from the total likelihood maximization have been summarized in Table 1. With these estimates, the posterior probability that record i from dataset A and record j from dataset B belong to each other was calculated. A posterior probability of>80% was considered high enough to assume that both deliveries belonged to the same mother. With this arbitrary threshold, we were able to identify 304,130 potential record pairs (Table 2).
      SUPPLEMENTARY TABLE 1Estimated parameters that were derived from the maximumization of the total likelihood function
      Linkage variableParameters likelihood
      Logit, μLogit, υ
      v14.66−8.8
      v22.16−8.3
      v31.19−7.50
      π−13.80
      Schaaf. Recurrence risk of preterm singleton birth after twin delivery. Am J Obstet Gynecol 2012.
      SUPPLEMENTARY TABLE 2Number of patterns with the highest posterior probability
      Outcome comparisonFrequencyP (match)
      v1v2v3
      01125262.04
      10128952.49
      Missing113.83→ Threshold for linkage
      11042963.84
      1Missing1139359.90
      11Missing315.96
      111121490.99
      Patterns with a posterior probability >0.80 were considered as a linkage.
      Schaaf. Recurrence risk of preterm singleton birth after twin delivery. Am J Obstet Gynecol 2012.
      After the linkage, the dataset was checked. Some deliveries from dataset A were wrongly linked to deliveries from dataset B. We reduced the number of false-positive pairs of second and first deliveries with the following steps: (1) the exclusion of pairs in which deliveries from dataset A were linked to >1 delivery in dataset B and (2) the exclusion of pairs of deliveries in which the second delivery (from dataset A) happened <10 months after the first delivery (from dataset B).
      We validated our longitudinal linkage in a random sample of 200 deliveries within the Academic Medical Center Amsterdam and found that the calculated posterior probability is similar to the actual chance of a correct linkage of 2 subsequent deliveries.
      After these steps, we were able to identify 272,551 pairs of first and second deliveries. Table 3 shows that basic baseline and pregnancy characteristics are comparable between the original and linked datasets. Differences in baseline characteristics were assessed with the Student t test for normally distributed continuous variables and the chi-square test for categoric variables.
      SUPPLEMENTARY TABLE 3Comparison of baseline characteristics of original dataset and linked dataset for first and second deliveries
      CharacteristicFirst deliverySecond delivery
      Original dataset (n = 667,053)Linked dataset (n = 272,551)Original dataset (n = 509,559)Linked dataset (n = 272,551)
      Gestational age, wk
      Data are given as mean ± SD;
      39.1 ± 2.339.2 ± 2.2
      P < .05 compared with the original dataset.
      39.3 ± 1.939.3 ± 1.8
      Birthweight, g
      Data are given as mean ± SD;
      3315 ± 6153346 ± 608
      P < .05 compared with the original dataset.
      3510 ± 5813539 ± 573
      P < .05 compared with the original dataset.
      Maternal age, y
      Data are given as mean ± SD;
      28.8 ± 4.828.6 ± 4.2
      P < .05 compared with the original dataset.
      31.2 ± 4.331.2 ± 4.2
      White maternal ethnicity, n (%)574,809 (86.2%)242,907 (89.1%)
      P < .05 compared with the original dataset.
      437,832 (85.9%)242,801 (89.1%)
      P < .05 compared with the original dataset.
      Twin pregnancy, n (%)14,138 (2.1%)4346 (1.6%)
      P < .05 compared with the original dataset.
      10,065 (2.0%)5383 (2.0%)
      Preterm birth <37 weeks, n (%)58,791 (8.8)22,080 (8.1%)
      P < .05 compared with the original dataset.
      26,162 (5.1%)13,185 (4.8%)
      P < .05 compared with the original dataset.
      Very preterm birth <32 weeks, n (%)9772 (1.5)3760 (1.4%)
      P < .05 compared with the original dataset.
      3911 (0.8%)1757 (0.6%)
      P < .05 compared with the original dataset.
      Congenital malformations, n (%)16,983 (2.6)6777 (2.5%)11,104 (2.2%)6120 (2.3%)
      Perinatal deaths, n (%)6203 (9.3)3664 (13.4%)
      P < .05 compared with the original dataset.
      3448 (6.8%)1666 (6.1%)
      P < .05 compared with the original dataset.
      Schaaf. Recurrence risk of preterm singleton birth after twin delivery. Am J Obstet Gynecol 2012.
      a Data are given as mean ± SD;
      b P < .05 compared with the original dataset.
      Because of the large datasets, we found several statistically significant differences between the linked and original datasets. However, only the differences in multiple pregnancy and perinatal mortality rates in the first pregnancy appear to be clinically significant differences.

      References

        • Chauhan S.P.
        • Scardo J.A.
        • Hayes E.
        • Abuhamad A.Z.
        • Berghella V.
        Twins: prevalence, problems, and preterm births.
        Am J Obstet Gynecol. 2010; 203: 305-315
        • Imaizumi Y.
        Trends of twinning rates in ten countries, 1972-1996.
        Acta Genet Med Gemellol (Roma). 1997; 46: 209-218
        • Eriksson A.W.
        • Fellman J.
        Temporal trends in the rates of multiple maternities in England and Wales.
        Twin Res Hum Genet. 2007; 10: 626-632
        • Martin J.A.
        • Hamilton B.E.
        • Ventura S.J.
        • et al.
        Births: final data for 2009.
        Natl Vital Stat Rep. 2011; 60: 1-70
        • Ananth C.V.
        • Vintzileos A.M.
        Epidemiology of preterm birth and its clinical subtypes.
        J Matern Fetal Neonatal Med. 2006; 19: 773-782
        • Wen S.W.
        • Smith G.
        • Yang Q.
        • Walker M.
        Epidemiology of preterm birth and neonatal outcome.
        Semin Fetal Neonatal Med. 2004; 9: 429-435
        • Slattery M.M.
        • Morrison J.J.
        Preterm delivery.
        Lancet. 2002; 360: 1489-1497
        • Adams M.M.
        • Elam-Evans L.D.
        • Wilson H.G.
        • Gilbertz D.A.
        Rates of and factors associated with recurrence of preterm delivery.
        JAMA. 2000; 283: 1591-1596
        • Goldenberg R.L.
        • Iams J.D.
        • Mercer B.M.
        • et al.
        The preterm prediction study: the value of new vs standard risk factors in predicting early and all spontaneous preterm births: NICHD MFMU Network.
        Am J Public Health. 1998; 88: 233-238
        • Bloom S.L.
        • Yost N.P.
        • McIntire D.D.
        • Leveno K.J.
        Recurrence of preterm birth in singleton and twin pregnancies.
        Obstet Gynecol. 2001; 98: 379-385
        • Facco F.L.
        • Nash K.
        • Grobman W.A.
        Are women who have had a preterm twin delivery at greater risk of preterm birth in a subsequent singleton pregnancy?.
        Am J Obstet Gynecol. 2007; 197: 253
        • Menard M.K.
        • Newman R.B.
        • Keenan A.
        • Ebeling M.
        Prognostic significance of prior preterm twin delivery on subsequent singleton pregnancy.
        Am J Obstet Gynecol. 1996; 174: 1429-1432
        • Rydhstroem H.
        Gestational duration in the pregnancy after a preterm twin delivery.
        Am J Obstet Gynecol. 1998; 178: 136-139
        • Rafael T.J.
        • Hoffman M.K.
        • Leiby B.E.
        • Berghella V.
        Gestational age of previous twin preterm birth as a predictor for subsequent singleton preterm birth.
        Am J Obstet Gynecol. 2012; 206: 156
        • Meray N.
        • Reitsma J.B.
        • Ravelli A.C.
        • Bonsel G.J.
        Probabilistic record linkage is a valid and transparent tool to combine databases without a patient identification number.
        J Clin Epidemiol. 2007; 60: 883-891
        • Tromp M.
        • Ravelli A.C.
        • Meray N.
        • Reitsma J.B.
        • Bonsel G.J.
        An efficient validation method of probabilistic record linkage including readmissions and twins.
        Methods Inf Med. 2008; 47: 356-363
        • Schaaf J.
        • Mol B.
        • Abu-Hanna A.
        • Ravelli A.
        Trends in preterm birth: singleton and multiple pregnancies in the Netherlands, 2000-2007.
        BJOG. 2011; 118: 1196-1204
        • Zeitlin J.
        • Di L.D.
        • Blondel B.
        • et al.
        Variability in caesarean section rates for very preterm births at 28-31 weeks of gestation in 10 European regions: results of the MOSAIC project.
        Eur J Obstet Gynecol Reprod Biol. 2010; 149: 147-152
        • Van der Ham D.P.
        • Nijhuis J.G.
        • Mol B.W.
        • et al.
        Induction of labour versus expectant management in women with preterm prelabour rupture of membranes between 34 and 37 weeks (the PPROMEXIL-trial).
        BMC Pregnancy Childbirth. 2007; 7: 11
        • Meis P.J.
        • Klebanoff M.
        • Thom E.
        • et al.
        Prevention of recurrent preterm delivery by 17 alpha-hydroxyprogesterone caproate.
        N Engl J Med. 2003; 348: 2379-2385