Recent studies have reported lower fetal loss rates after amniocentesis compared with the loss rate of 1 in 200 used by most practitioners in counseling women before the procedure.1,2 The fetal loss rate after amniocentesis reported from the First and Second Trimester Evaluation of Risk (FASTER) trial was 0.06%, or 1 in 1,600,1 and a recent report using data from the state of California in women having amniocentesis for abnormal serum screen was 0.46% compared with 0.53% in those who did not have amniocentesis.2 The estimated loss rate of 1 in 200 is based on recommendations by the Centers for Disease Control and Prevention (CDC) and endorsed by the American College of Obstetricians and Gynecologists (ACOG).3,4 The loss rates reported in these more recent studies are different from those reported from previous reports.5,6
With the publication of the new ACOG recommendations for screening for aneuploidy,7 we expect in the future more women than in past years will request invasive procedures, making the availability of precise estimates of the procedure-related loss rate during counseling crucial. Instead of using the generic rate recommended by the CDC and ACOG or using the lower rate reported by the FASTER consortium,1,3,4 we sought to estimate our institution-specific fetal loss rate after amniocentesis.
MATERIALS AND METHODS
To estimate the fetal loss rate following amniocentesis at Washington University School of Medicine, we searched our prenatal diagnosis database for all pregnant women presenting for care between 1990 and 2006. The study was approved by the Washington University School of Medicine human studies review board. Since creation of the database in 1988, outcome data for all women seen for prenatal diagnosis have been obtained by dedicated pregnancy outcome coordinators. After delivery, each patient who is seen in our prenatal diagnosis center is asked to return a standardized form requesting the pregnancy outcome. When a form is not returned within 4 weeks of the expected date of delivery, the patient receives a phone call from the coordinator. In cases where the patient cannot be contacted, the referring physician is contacted for the outcome information. For patients delivering in our health care system, outcome data were extracted from our perinatal computerized database. The median follow-up period was 12 weeks (range: 1 day to 24 weeks). The standardized follow-up sheet includes details about pregnancy complications, delivery indications, and neonatal outcomes including chromosomal and structural abnormalities. Maternal demographics, past obstetric history, indications for the ultrasound visit, findings from the ultrasound examination or any testing performed, and outcome of the pregnancy were all collected and stored in the database. All amniocentesis procedures were performed under ultrasound guidance by trained perinatologists and obstetricians, with further specialist training in medical genetics or by Fellows in maternal–fetal medicine under the supervision of the attending physicians. No procedure was performed by residents. Procedures were performed with a 22-gauge needle, and in the majority of cases 20 mL of amniotic fluid were collected.
We compared the fetal loss rate in women who had amniocentesis between 15 and 22 weeks of gestation to that of all women who did not have any invasive procedure and had a live fetus documented on ultrasound examination between 15 and 22 weeks. Only singleton gestations were included. Those having chorionic villus sampling or pregnancy termination before 15 weeks were excluded from this analysis. Fetal loss rate was defined as any loss before 24 weeks. Elective terminations of pregnancy were not considered as pregnancy loss in this analysis. Patients with miscarriages or intrauterine fetal death before 24 weeks were defined as “spontaneous” fetal losses. However, deliveries resulting in fetal death after induction of labor for ruptured membranes or oligohydramnios before 24 weeks were included in our final definition of a fetal loss at less than 24 weeks. Statistical methods, including univariable and logistic regression analysis, were used to identify and adjust for potential confounders between the groups. Chi-square, Fisher exact, and Student t tests were used for categorical and continuous variables, as appropriate. Log-rank test was used to compare the timing of miscarriage between the amniocentesis group and those without an invasive procedure. All statistical analyses were performed using Stata 9.1 (SE) (StataCorp, College Station, TX).
Over the 16-year period, the database collected information on 64,227 women seen before 24 weeks. We excluded 5,477 women who had chorionic villus sampling and 314 who had spontaneous losses before 15 weeks. Therefore, 58,436 women in the database met the inclusion criteria. Of these, 12,916 women had amniocentesis and 45,520 did not. Complete outcome data were available for 51,577 (88%), 11,746 of 12,916 (91%) in the amniocentesis group and 39,821 (87%) in those who did not have amniocentesis. Exact gestational age at delivery was available in 51,289 of 51,577 (99%) women in the cohort, 11,695 of 11,746 (99.6%) in the amniocentesis group and 39,594 of 39,821 (99.4%) in the group that had no procedure, respectively.
The overall characteristics of the two groups are shown in Table 1. The women in the amniocentesis group were seen at a significantly earlier mean gestational age than those in the control group. The women in the amniocentesis group were significantly older, had more smokers, and more women exposed to alcohol compared with those in the group that had no procedure. Preterm delivery (less than 32 weeks) was significantly higher in the group that had no procedure compared with the amniocentesis group. Other significant differences between the two groups are shown in Table 1. In the overall study population, there were 149 (0.29%) pregnancy losses that were not the result of labor induction or active management, 47 (0.4%) in those who had amniocentesis compared with 102 (0.26%) in the group without amniocentesis (relative risk 1.6, 95% confidence interval [CI] 1.1–2.2). Assuming the spontaneous fetal loss rate in the group without an invasive procedure to be the background loss rate, the pregnancy loss rate attributable to amniocentesis (excluding those induced) is 0.14% (95% CI 0.019–0.26%), 1 in 714. The overall fetal loss rate (including losses due to labor induction for preterm premature rupture of membranes and oligohydramnios) at less than 24 weeks for the entire cohort was 0.87% (448 of 51,289), 113 of 11,695 (0.97%) in the group that had amniocentesis and 335 of 39,594 (0.84%) in those that had no procedure, respectively. The difference of 0.13% (95% CI –0.07 to 0.2%) was not statistically significant, P=.18. The incidence of ruptured membranes occurring before 24 weeks was 21 of 113 (19%) in the amniocentesis group compared with 101 of 335 (30%) in the group that had no invasive procedure (P<.02). After adjusting for confounding variables identified from the univariable analyses, there was no significant difference in total fetal loss rates between those who had amniocentesis or those who did not have an invasive procedure (Table 2). In the subgroup analyses, the only subgroup that had significantly different loss rates was in the group that had a normal second-trimester serum aneuploidy screen, with a loss rate of 0.97% after amniocentesis compared with 0.80% in the same subgroup with no procedure. The difference remained statistically significant after adjusting for variables that were significant in the univariable analysis (Table 2).
The median gestational ages at fetal loss were 20.6 (range: 15.7–23.9) and 21.5 weeks (range: 16.1–23.9 weeks) in the amniocentesis and no-procedure groups, respectively. Using these data, we evaluated the cumulative risk of fetal loss by the number of weeks from prenatal diagnosis using a Kaplan-Meier plot (Fig. 1). There was a significantly higher risk for fetal loss in the no-procedure group for the period between 15 and 24 weeks after prenatal diagnosis compared with the amniocentesis group (log rank P<.001). There did not appear to be any temporal clustering in the loss rate closer to the time of the procedure in the amniocentesis group. The fetal loss rates in the two groups remained stable over the study period, with ranges from 0.8–1.4% and 0.6–1.2% in the amniocentesis group compared with the no-procedure group, respectively (details available from authors).
To determine if the differences in mean gestational age at prenatal diagnosis between the groups may have influenced the findings, we conducted a nested case–control study from the entire cohort. Each case that had an amniocentesis was matched to three cases that had no procedure, matching for gestational age (±6 days). The cases and their matched controls were chosen consecutively from the database. Due to the smaller numbers of controls with no procedures seen between 15 weeks and 17.9 weeks, we used a 1:1 matching ratio at those gestational ages. Due to this limitation in the early gestational ages, we could find a match for only 11,196 of the 11,695 (96%) of the amniocentesis group. Of 29,886 total subjects matched, there were 312 fetal losses (less than 24 weeks), 106 of 11,196 (0.94%) in the amniocentesis group and 206 of 18,690 (1.1%) in the matched control group, P=.11. Figure 2 shows the survival curve comparing the likelihood of surviving up to 24 weeks in the cases compared with the matched controls. There was a statistically significant increase in the rate of fetal loss in the no-procedure group compared with the amniocentesis group (log rank test P=.001). A similar pattern was seen for delivery at less than 28 weeks (data available from authors).
We evaluated potential variables that may be associated with fetal loss after amniocentesis. Among the women having amniocentesis, 202 had two needle insertions and one had three. The significant factors associated with a fetal loss after amniocentesis were the presence of a fetal structural abnormality, a history of a previous miscarriage, and a dark-brownish discoloration of the amniotic fluid (Table 3). Other discolorations of the amniotic fluid, including blood staining, were not associated with increased loss (data available from authors).
Assuming an alpha of 0.5, the study had 88% power to detect a 50% increase in fetal loss from amniocentesis above the background risk of 0.84% in the group not having amniocentesis. Table 4 summarizes the findings of published studies comparing amniocentesis-related loss rates with control groups that received no intervention.1,2,5,8–13 The differences in definitions of fetal loss and some of the limitations of the studies are depicted.
This study found no significant difference in fetal loss rates at less than 24 weeks between pregnancies after amniocentesis when compared with those without any invasive procedure. Our findings of a loss rate of 1:769 (95% CI 1:500 to 1:1429) adds to the growing body of evidence supporting the safety of amniocentesis.1,2,13 Eddleman et al1 reported a procedure-related fetal loss rate after second-trimester amniocentesis of 0.06% or 1:1600. Caughey et al13 reported a loss rate of 0.46 in a retrospective cohort study of 30,893 women undergoing amniocentesis. The procedure-related loss rate in the last 5 years of the latter study was 0.27%, or 1 in 370. In another retrospective study limited to women with abnormal serum screen from California, the rate of miscarriage after amniocentesis was 0.46%, which was lower than the 0.53% seen in women who did not have amniocentesis.2 The differences between studies on fetal loss after amniocentesis that included a control group that had no procedure are listed in Table 4. The inclusion criteria and definitions of fetal loss are variable. All of the listed studies did not explicitly define fetal loss as we did in the present study.
Our study has the added advantage of including a large number of women who had no invasive procedure as a comparison group. The fetal loss rate in both the amniocentesis and control groups from our study is, however, lower than that reported from most of the previous studies addressing this subject.5,6,14 One reason for this may be the inclusion of only women seen from 15 weeks onwards. Because this inclusion criterion was applied equally to both groups, we believe our conclusions to be reasonable and valid. Using high-resolution ultrasound guidance and limiting our needle size for the procedure to 22 gauge may also have contributed to the lower loss rate in our study (Alfirevic Z, Tabor A. Pregnancy loss rates after midtrimester amniocentesis [letter]. Obstet Gynecol 2007;109:1203–4).
When we considered the risk of fetal loss from the time of prenatal diagnosis, we found a higher risk in the group that had no procedure compared with the amniocentesis group. This higher risk persisted throughout the period between 15 weeks and 24 weeks. Although this finding does not imply that the amniocentesis procedure is without risk, it suggests that the risk is not significantly different from the background risk for spontaneous miscarriage. From a public health standpoint, a 0.13% procedure-related loss rate may become clinically significant when applied to a large population, particularly if the procedure in question is (or is anticipated to be) performed with increasing frequency.
The only significant factors associated with fetal loss in the analysis limited to the amniocentesis group were a brownish discoloration of the amniotic fluid, the presence of a fetal abnormality, and a history of miscarriage. This finding of an association between brownish amniotic fluid color and fetal loss is similar to that reported in previous studies.5,15 We, however, believe the discoloration to be secondary to pre-existing intraamniotic bleeding and subsequent amniorrhexis from enzymatic digestion of membranes, rather than from the amniocentesis procedure per se. Unlike previous reports, we found no association in this study between the number of needle insertions, transplacental passage, maternal age, or pregnancy termination and fetal loss after amniocentesis.5,16 In the subgroup analysis, the only group with a significantly different fetal loss rate compared with the group that had no procedure was in the group of women undergoing amniocentesis with normal serum screen for aneuploidy. Despite adjusting for other potential confounders including maternal age and the presence of a fetal anomaly, the difference remained statistically significant. The mechanism for this finding is unclear, but the loss rate in this subgroup of 0.17%, or 1:588, is still lower than the 1:200 suggested by the CDC.
Our study is not without limitations. These include the retrospective design and the possibility of selective follow-up in the amniocentesis group. The latter should lead to a detection of a higher loss rate in the amniocentesis group; therefore, our findings may be a conservative estimate if this is truly the case. The follow-up schedule described has been in place in our center for many years and was not designed specifically for the current study. Therefore, a differential follow-up is an unlikely reason for our findings. Although a randomized control trial would be ideal in determining the fetal loss rate associated with amniocentesis, we agree with Eddleman et al1 that such a study will be impractical with the present prenatal diagnostic practice patterns. The only randomized trial addressing this subject was performed between 1982 and 1984 and enrolled only low-risk women under 35 years of age. Those investigators reported a spontaneous loss rate of 1.7% in the amniocentesis group compared with 1% in the control group.5 In addition, an 18-gauge needle was used for the amniocentesis in this study by Tabor et al. In our cohort, documentation of the exact timing of the fetal loss was present in over 90% of those who had a fetal loss. We agree with the opinion of Nicolaides that this information is important in reviewing the accuracy of studies addressing the loss rate from procedures (Nicolaides K. Pregnancy loss rates after midtrimester amniocentesis [letter]. Obstet Gynecol 2007;109:780).
A recent systematic review has suggested matching for gestational age between the control group and those undergoing amniocentesis as a solution to this potential for bias.17 However, matched case-control studies have the disadvantages of introducing potential selection bias into the study and of not providing reliable incidence rates for some of the outcomes being studied because they are already fixed by the matching process.18 Because the women in the amniocentesis group were enrolled at a significantly earlier gestational age than those in the control group, we believed that this could bias the results in favor of a higher loss rate in the amniocentesis group compared with the control group, particularly because previous studies have shown that the higher miscarriage rates occur in the earlier gestational ages than later.19 We explored this possibility by performing a nested case-control study by matching for gestational age at prenatal diagnosis. The results of this analysis revealed a higher loss rate in the group that had no procedure compared with the amniocentesis group. The matched analysis was limited by fewer numbers of controls in the earlier gestational ages between 15 and 17.9 weeks. This study is not the first to report a lower fetal loss in the amniocentesis group compared with those who had no procedure.2 Similarly, in the adjusted model from the report by Eddleman et al,1 the odds of pregnancy loss were lower in the women who had amniocentesis compared with those who did not. The suggested reason for this finding is that, given a high association between chromosomal abnormalities and fetal loss and since the majority of pregnancies with such abnormalities are terminated after amniocentesis, a higher loss rate is seen in the group that had no procedure due to a relatively higher proportion of pregnancies with aneuploidy in those who had no invasive procedure. The difficulty with equal comparison is further confounded by the lack of chromosomal analysis in a high proportion of spontaneous fetal losses. One possible mechanism for adjusting for this potential bias is to estimate the expected rates of different chromosomal abnormalities in the control group and the expected loss rates in such pregnancies—for example, 40% of Down syndrome pregnancies will spontaneously miscarry by 24 weeks19—and, finally, reduce the loss rate in the no-procedure group by this factor. Because we have no reliable data about the incidence of chromosomal abnormalities and their loss rates in women who had no procedure in our population, we cannot perform such adjustments.
In conclusion, our study found the fetal loss rate after amniocentesis to be lower than the 1 in 200 risk suggested by the CDC and ACOG,3,4 and in our cohort is estimated to be 1 in 769. We also found that the total fetal loss rate after amniocentesis is not significantly higher than the risk in women who had no invasive procedure. This information will be useful for counseling women requesting amniocentesis. Other centers might consider validating these findings before adopting the reported rate for their unit, particularly if the center or patient population appears unlike ours.
1. Eddleman KA, Malone FD, Sullivan L, Dukes K, Berkowitz RL, Kharbutli Y, et al. Pregnancy loss rates after midtrimester amniocentesis. Obstet Gynecol 2006;108:1067–72.
2. Towner D, Currier RJ, Lorey FW, Cunningham GC, Greve LC. Miscarriage risk from amniocentesis performed for abnormal maternal serum screening. Am J Obstet Gynecol 2007;196:608.e1–5.
3. Chorionic villus sampling and amniocentesis: recommendations for prenatal counseling. Centers for Disease Control and Prevention. MMWR Recomm Rep 1995;44 (RR-9):1–12.
4. American College of Obstetricians and Gynecologists. Prenatal diagnosis of fetal chromosomal abnormalities. ACOG Practice Bulletin 27. Washington (DC): ACOG; 2001.
5. Tabor A, Philip J, Madsen M, Bang J, Obel EB, Norgaard-Pedersen B. Randomised controlled trial of genetic amniocentesis in 4606 low-risk women. Lancet 1986;1:1287–93.
6. Seeds JW. Diagnostic mid trimester amniocentesis: how safe? Am J Obstet Gynecol 2004;191:607–15.
7. Screening for fetal chromosomal abnormalities. ACOG Practice Bulletin No. 77. American College of Obstetricians and Gynecologists. Obstet Gynecol 2007;109:217–27.
8. Farahani G, Goldman MA, Davis JG, Kardon NB, Mohandes E, Pek H, et al. Use of the ultrasound aspiration transducer in midtrimester amniocentesis J Reprod Med 1984;29:227–31.
9. Andreasen E, Kristoffersen K. Incidence of spontaneous abortion after amniocentesis: influence of placental localization and past obstetric and gynecologic history Am J Perinatol 1989;6:268–73.
10. Tongsong T, Wanapirak C, Sirivatanapa P, Piyamongkol W, Sirichotiyakul S, Yampochai A. Amniocentesis-related fetal loss: a cohort study Obstet Gynecol 1998;92:64–7.
11. Antsaklis A, Papantoniou N, Xygakis A, Mesogitis S, Tzortzis E, Michalas S. Genetic amniocentesis in women 20–34 years old: associated risks Prenat Diagn 2000;20:247–50.
12. Muller F, Thibaud D, Poloce F, Gelineau MC, Bernard M, Brochet C, et al. Risk of amniocentesis in women screened positive for Down syndrome with second trimester maternal serum markers. Prenat Diagn 2002;22:1036–9.
13. Caughey AB, Hopkins LM, Norton ME. Chorionic villus sampling compared with amniocentesis and the difference in the rate of pregnancy loss Obstet Gynecol 2006;108:612–6.
14. Randomised trial to assess safety and fetal outcome of early and midtrimester amniocentesis. The Canadian Early and Mid-trimester Amniocentesis Trial (CEMAT) Group. Lancet 1998;351:242–7.
15. Golbus MS, Stephens JD, Cann HM, Mann J, Hensleigh PA. Rh isoimmunization following genetic amniocentesis. Prenat Diagn 1982;2:149–56.
16. Papantoniou NE, Daskalakis GJ, Tziotis JG, Kitmirides SJ, Mesogitis SA, Antsaklis AJ. Risk factors predisposing to fetal loss following a second trimester amniocentesis. BJOG 2001;108:1053–6.
17. Mujezinovic F, Alfirevic Z. Procedure-related complications of amniocentesis and chorionic villus sampling: a systematic review. Obstet Gynecol 2007;110:687–94.
18. Rothman KJ, Greenland S. Matching. In: Rothman KJ, Greenland S. Modern epidemiology. 2nd ed. Philadelphia (PA): Lippincott Williams & Williams; 1998. p. 147–61.
19. Won RH, Currier RJ, Lorey F, Towner DR. The timing of demise in fetuses with trisomy 21 and trisomy 18. Prenat Diagn 2005;25:608–11.