Tong, Stephen PhD; Ngian, Gene-Lyn MBBS; Onwude, Joseph L. MBChB; Permezel, Michael MD; Saglam, Burcu PhD; Hay, Sarah RN; Konje, Justin C. PhD; Marczylo, Tim H. PhD; Fleming, Gabrielle RN; Walker, Susan P. MD; Lappas, Martha PhD
Miscarriage is the most common complication of pregnancy.1,2 Currently, there are no treatments that can prevent most cases of sporadic miscarriage and no accurate predictive tests. Although approximately 50% are associated with chromosomal errors,2 the remainder are likely to be attributable to suboptimal implantation and therefore are potentially salvageable.
A predictive test for miscarriage would be useful for counseling purposes, allowing clinicians to reassure women very early in their pregnancy. Second, it would enable researchers or clinicians to target administration of potential therapies to women at high risk for miscarriage. In this regard, promising leads are emerging for the treatment of miscarriage. For instance, pravastatin has been shown to rescue embryos in a miscarriage animal model,3 and a number of newer-generation anticoagulant agents4,5 potentially merit evaluation.
The current literature on miscarriage prediction has mainly investigated the ability of putative markers to predict miscarriage among symptomatic women.6–10 Hence, samples were only obtained once women had presented with symptoms of miscarriage (abdominal pain and bleeding). In this setting, the pathophysiological process will already be advanced (ie, hemorrhage into subchorionic spaces) to the point at which it may be arguably very difficult for any therapeutic agent to rescue the pregnancy. Potentially, a particularly useful biomarker would be one that can predict miscarriage among asymptomatic women with a viable fetus. It could flag pregnancies at high risk for miscarriage at an otherwise early stage of the pathophysiological process, when potential therapeutic interventions may be afforded an opportunity to salvage these pregnancies.
Using retrospective banked samples, we previously reported that maternal serum levels of pregnancy-associated plasma protein-A (PAPP-A)11 and (macrophage inhibitory cytokine-1)1 at 7–13 weeks of gestation were very significantly decreased in those destined to miscarry at just 11% and 31% of normal levels, respectively. These women, enrolling in a Down syndrome screening study, were asymptomatic at the time of recruitment. Although these were highly encouraging results, a limitation of both these retrospective reports was that we did not obtain evidence of fetal cardiac activity on the day of blood sampling. Therefore, the potential predictive value of these analytes required validation in a prospective cohort study.
Therefore, we implemented the Mercy Early Pregnancy Study, a large prospective collection of serum samples from asymptomatic women presenting in the middle of the first trimester for their first prenatal visit. Crucially, we obtained confirmation of fetal cardiac activity the same day we collected the blood sample. The primary purpose was to validate whether serum macrophage inhibitory cytokine-1 and PAPP-A level could predict miscarriage and, if so, to determine their biomarker performance characteristics.
Habayeb et al10 reported that among women with threatened miscarriage, increased plasma anandamide levels were strongly associated with subsequent miscarriage. Hence, we also measured anandamide. Finally, we assayed serum β-human chorionic gonadotropin (hCG) because it is a commonly used clinical biomarker for a number of early pregnancy complications.
MATERIALS AND METHODS
Between 2004 and 2008, we implemented the Mercy Early Pregnancy Study. This was a prospective cohort study recruiting asymptomatic (no vaginal bleeding or abdominal or pelvic pain) women attending their first prenatal clinic visit between 6 0/7 and 12 6/7 weeks of gestation. We asked the attending clinician to confirm fetal viability by recording the presence of nonbradycardic fetal cardiac activity by ultrasound examination at the clinic on the day of sample collection. Thus, those with a miscarriage diagnosed at the first prenatal appointment (ie, no fetal cardiac activity) were not recruited. Those with a multiple pregnancy diagnosed were also excluded.
All women were recruited at Mercy Hospital for Women, which is a large tertiary care hospital in Melbourne, Victoria, providing care for more than 5,000 pregnancies per year. Part-time research midwives were employed for this study, meaning we were able to recruit at a rate of approximately 10–15 patients per week.
We recorded baseline clinical characteristics and delivery outcomes. Gestational age was derived from first trimester ultrasound scans. Miscarriage was defined as a pregnancy loss at less than 20 weeks of gestation. This study was approved by the Mercy Hospital for Women Ethics Committee (approval code R03/23), and all women gave informed written consent to participate in the study.
The overall aim was to verify the ability of four serum analytes to predict miscarriage: macrophage inhibitory cytokine-1, PAPP-A, anandamide, and β-hCG. We used a two-step approach in which we first screened these four analytes in all those in the case group (n=21) and a subgroup randomly selected (n=115) from the 761 available women who did not miscarry (ongoing pregnancy group). This subgroup of 115 who had an ongoing pregnancy were randomly selected but were stratified according to gestation when samples were collected (15–20 samples from each gestational week). Analytes that showed a significant difference (PAPP-A and macrophage inhibitory cytokine-1) between those in the case group and those in the ongoing pregnancy subgroup were then assayed in the entire cohort.
Approximately 8 mL of venous blood was collected into vacuette serum clot separator tubes without additives. The tubes were centrifuged at 3,000 g for 10 minutes, and the serum was collected and stored at −80°C until analysis. All samples were processed promptly and frozen within hours of sample collection.
Serum samples were thawed and assayed in batch to measure the analytes of interest by investigators blinded to clinical outcomes. Standard enzyme-linked immunosorbent assay kits for PAPP-A (limit of detection 0.133 mg/mL), macrophage inhibitory cytokine-1 (limit of detection 8 pg/mL), and β-hCG (limit of detection 5 international units/L) were used according to the manufacturer's instructions. For all assays, manufacturer specifications indicate that intra-assay and interassay coefficients of variation are less than 10%. Anandamide levels were measured in serum using a previously published ultra-high-performance liquid chromatography electrospray tandem mass spectroscopy method12 following sample clean-up by solid-phase extraction using Oasis HLB cartridges (1 mL).13
For statistical comparisons between two groups, we used t tests to compare continuous data that were normally distributed and Mann-Whitney U to compare data that were not normally distributed. Normality of data was assessed with a normal probability plot. The distribution was accepted to be normal if the analyte data and the normal quantiles (a measure of goodness of fit) were both approximately on a straight line. Kruskal-Wallis test was used to examine whether analyte concentrations changed across gestation. The χ2 test was used to compare categorical data.
To allow comparison of levels between the women who had miscarriages and those who had an ongoing pregnancy, the normal median levels of each analyte were calculated for each completed week of pregnancy from those who had ongoing pregnancies. Each data point was then expressed as a multiple of the normal median (multiples of the median [MOM]).
Given macrophage inhibitory cytokine-1 and PAPP-A were significantly different between those who miscarried and those who had ongoing pregnancies, we estimated their degree of independence by measuring their correlation coefficient. Because this was minimal (r=−0.36), we treated them as independent analytes. We performed multiple logistic regressions to calculate the probability of predicting a miscarriage with either macrophage inhibitory cytokine-1 alone or PAPP-A alone.
We subsequently combined the two analytes when we performed receiver-operated characteristics curve analysis to determine the cut-off values for macrophage inhibitory cytokine-1 and PAPP-A to predict miscarriage. We determine sensitivities, specificities, positive and negative predictive values, and likelihood ratios of these predictive biomarkers. The χ2 test was used to compare the areas under the receiver-operated characteristics curves.
Between 2004 and 2008, we prospectively collected 1,044 serum samples from asymptomatic women at 6 0/7 to 12 6/7 weeks of gestation. We obtained ultrasound evidence of fetal cardiac activity on the day of blood collection. Given that all cases of subsequent miscarriage occurred in women from whom samples were obtained between 6 0/7 and 10 6/7 weeks of gestation, we restricted analysis to samples collected across this same gestational age range (thus, excluding samples collected at 11 0/7 and 12 6/7 weeks of gestation for this study). A total cohort of 21 women who had miscarriages and 761 women with ongoing pregnancies therefore was available for this study.
The mean (±standard error of the mean) maternal age among those who miscarried was 32.2 (±1.4) years, which is no different from those who had ongoing pregnancies (31.1±0.2; P=.269). At the time of blood sample collection, 14% (3 of 21) of those who miscarried were primiparous women compared with 32% (246 of 761) of those who had ongoing pregnancies (P=.08). The mean (±standard deviation) gestational age at delivery for the group who had ongoing pregnancies was 39.2 (±1.8) weeks and the mean (±standard deviation) birth weight was 3,395 g (±554); 55% (419 of 761) had a spontaneous vaginal birth, 13% (99/761) required an instrumental delivery, and 32% (243/761) had a caesarean delivery. All those with ongoing pregnancies had a live birth.
Table 1 details the clinical history of the 21 women who miscarried. Given that it is not possible to determine the exact moment of fetal demise, we have presented the gestational ages when the women presented with symptoms of miscarriage and when the formal ultrasound examination was performed confirming a miscarriage. In many instances, there was a considerable period between sample collection and the ultrasound diagnosis of miscarriage: 1–3 weeks in nine women, 4–7 weeks in six women, 8 or more weeks in four women, and undetermined in two women (Table 1).
To rationalize use of samples and reagents, we first screened the four analytes in the 21 women who had a miscarriage and compared the results to a subgroup of 115 women with ongoing pregnancies. This subgroup was randomly selected from the 761 available women with ongoing pregnancies. Only analytes significantly different between those who miscarried and those in the subgroup with ongoing pregnancies were then assayed in the entire cohort. To correct for changes in analyte levels across gestation, the data were expressed as multiples of the normal gestational median (MOM).
The median MOM for serum β-hCG (25th–75th percentiles) among those who miscarried was 0.99 (0.46–1.86) and not different from control group individuals in the subgroup with ongoing pregnancies (1.00 [0.53–1.81]; P=.97; Fig. 1A). Median MOM for serum anandamide among those who miscarried was 1.07 (0.87–1.19), which was also not different from that of the subgroup with ongoing pregnancies (median MOM 1.00 [0.81–1.23]; P=.82; Fig. 1B). Serum β-hCG levels among those with ongoing pregnancies varied significantly across 6–10 weeks of gestation (P<.005), peaking at 7 weeks and declining thereafter (Fig. 1C). Levels of serum anandamide did not significantly vary across 6–10 weeks of gestation among the subgroup with ongoing pregnancies (P=.25; Fig. 1D).
Given that the median MOM (25th–75th percentiles) for both PAPP-A (0.35 [0.21–0.63]) and macrophage inhibitory cytokine-1 (0.68 [0.39–1.02]) were significantly decreased among those who miscarried compared with the subgroup with ongoing pregnancies (P<.002 for both comparisons), we proceeded to measure both analytes in the entire cohort of 761 women with ongoing pregnancies. In this report, we present the detailed results (from the next paragraph) of PAPP-A and macrophage inhibitory cytokine-1 from the entire group with ongoing pregnancies rather than the subgroup. For quality control, we reassayed the samples obtained from those who miscarried in the second analysis and found no difference between the two analyses (Mann-Whitney U test; P>.05).
Concentrations of both serum PAPP-A and macrophage inhibitory cytokine-1 were significantly lower among women who subsequently miscarried (n=21) compared with the results from 761 participants who had ongoing pregnancies. For macrophage inhibitory cytokine-1, the median MOM (25th–75th percentiles) was 0.63 (0.33–0.88) among those who miscarried compared with 1.00 (0.76–1.29) in 761 women who had ongoing pregnancies (Fig. 2A). For PAPP-A, the median MOM (25th–75th percentiles) was 0.23 (0.12–0.48) among those who miscarried compared with 1.00 (0.46–1.72) in 753 women who had ongoing pregnancies (Fig. 2B; P<.001 for both hormones). Both analytes increased significantly over the course of 6–10 weeks of gestation among samples obtained from ongoing pregnancies (P<.001; Fig. 2C and D).
At 90% specificity, the sensitivities of PAPP-A and macrophage inhibitory cytokine-1 in predicting miscarriage were 44% and 57%, respectively (Table 2). However, a test combining these analytes identified 63% (sensitivity) of women destined to miscarry at the same fixed false-positive rate of 10% (90% specificity). In this combined test, those with a positive test result had a 6.6 likelihood of miscarriage (likelihood ratio).
In predicting pregnancies destined to miscarry, the area under the curve (±95% confidence interval) for the test generated by combining macrophage inhibitory cytokine-1 and PAPP-A results was 0.83 (0.72–0.93) compared with 0.63 (0.50–0.77) for PAPP-A alone and 0.73 (0.63–0.84) for macrophage inhibitory cytokine-1 alone (Fig. 3). However, the area under the curves for PAPP-A (P=.07) and macrophage inhibitory cytokine-1 (P=.32) alone were not significantly different from area under the curve generated by combining these analytes.
Miscarriage is the most common complication of pregnancy, affecting 10–15% of clinically recognized pregnancies. Although there is no therapeutic treatment to prevent most cases of miscarriage, there continues to be avid interest in identifying one.14,15 It therefore would be useful if a noninvasive clinical test were to be available to identify women at high risk for miscarriage at an early stage of pregnancy, before symptoms have occurred. Such a test might allow targeted evaluation of potential therapies, limiting the exposure to women at high risk for miscarriage. This is especially important in this setting given that potential treatments would need to be administered during the time of embryogenesis. A further clinical application of a predictive test would be to use it as a counseling tool when, potentially, anxious women may be reassured (or, alternately, warned they are at risk) at a very early stage of pregnancy.
In this study, we undertook a large prospective study in asymptomatic women presenting to the prenatal clinic at 6–10 weeks of gestation, validating putative predictive serum biomarkers of miscarriage and characterizing their diagnostic performance characteristics. Importantly, we obtained confirmation of fetal cardiac activity by ultrasound examination in all women at the time of sample collection. We were able to confirm that serum PAPP-A and macrophage inhibitory cytokine-1, but not β-hCG or anandamide, are significantly altered among women destined to miscarry. Combining serum PAPP-A and macrophage inhibitory cytokine-1 produced a potential predictive test that performs strongly.
Although the miscarriage risk of clinically recognized pregnancies is approximately 10–15%,16 the rate in our study was 2.7%. The likely explanation for the differences in rates is that a number of women would have miscarried before attending their first booked prenatal appointment. It is therefore likely that we would have had a larger proportion miscarrying if we recruited all participants at 6 weeks of gestation.
We believe our study has several strengths. It is a large prospective study with a strong a priori aim to validate the predictive performance of biomarkers previously reported.1,11 Second, ours is a study that specifically recruited asymptomatic women presenting in the middle of the first trimester with confirmation of fetal viability at the time of sample collection. We felt this design would be the most relevant approach to investigate putative biomarkers for sporadic miscarriage. Possibly because of relative ease of sample and data ascertainment, the current literature investigating potential blood-based biomarkers of miscarriage mainly have investigated women already symptomatic with a threatened miscarriage at the time of recruitment,6–10 or examined markers measured in samples obtained at 10–13 weeks of gestation (from Down syndrome screening).17
Endocannabinoids have an important role in early pregnancy.18,19 Wang et al20 previously reported low anandamide levels to be a requirement for successful implantation. We have previously reported in a threatened miscarriage cohort the presence of high levels of plasma anandamide in those who subsequently miscarry.10 Maccarone et al21 also showed reduced levels of fatty acid amide hydrolase (the main enzyme that degrades anandamide) in the blood of women destined to miscarry. Although we failed to confirm increased anandamide in our cohort, it is important to note serum was examined in this study, not plasma. In addition, samples were frozen before assay in the current study, whereas in our previous study anandamide was measured within 2 hours of sample collection.10 This difference in sample processing could have additionally affected the results. Therefore, it is still possible that anandamide concentrations measured in plasma may predict miscarriage, and further investigations of this biomarker are underway.
Pregnancy-associated plasma protein-A is likely to play a critical role in placental function. It is a protease of insulin-growth factor binding protein-4.22 By cleaving insulin-growth factor binding protein-4, PAPP-A may be increasing free and bioactive insulin-like growth factors I and II. Serum PAPP-A is already measured clinically as part of the first trimester screening for Down syndrome at 10–13 weeks of gestation. Furthermore, low PAPP-A MOM at 10–13 weeks of gestation is associated with a number of adverse pregnancy outcomes, including preeclampsia, fetal growth restriction, and even stillbirth,17,23–25 although the accuracy (sensitivity and specificity) is limited. We had previously reported very low PAPP-A levels in association with miscarriage11 using a retrospective biobank and we now have confirmed this prospectively.
Macrophage inhibitory cytokine-1 is a member of the transforming growth factor-β superfamily1 and previously has been shown to be a placental product that localizes to the syncytiotrophoblast and increases across the first trimester in serum.26 It is also expressed in decidua.27 Its functional role in early pregnancy is still poorly characterized, although recent data suggest it might be promoting an increase in a tolerogenic subtype of dendritic cells in the decidua.27 If so, then its presence may be important to support a successful pregnancy. In support of this contention, we had previously reported decreased concentrations in association with miscarriage using specimens accessed from a biobank.1 In this present study, we have validated this finding.
We believe our data have a number of implications. First, we have identified two placental products that are likely to have important biological roles in maintaining early pregnancy. As such, they merit further investigation and consideration in fundamental basic science studies exploring early human pregnancy. In particular, the role of macrophage inhibitory cytokine-1 in early pregnancy remains poorly understood.
Second, our data showing decreased serum PAPP-A and macrophage inhibitory cytokine-1 preceding miscarriage by weeks strongly contrast with serum β-hCG, for which levels were similar between women with miscarriages and women in a control group with no apparent trend. Serum β-hCG is commonly used as a biomarker for early pregnancy complications. This suggests serum PAPP-A and macrophage inhibitory cytokine-1 merit further evaluation in other situations of early pregnancy where it may have better diagnostic characteristics than β-hCG.
Our main conclusion is that measuring serum PAPP-A and macrophage inhibitory cytokine-1 in the middle of the first trimester may form the platform of a promising predictive test of miscarriage. Potentially, our test can identify 63% (sensitivity of the combined test) of women who will ultimately miscarry, with a 10% false-positive rate (90% specificity). Arguably, such diagnostic performance characteristics already may be adequate as a screening tool to target potential therapies to women at high risk for pregnancy. However, it would be desirable to further enhance the sensitivity of this test, which may be achievable by identifying additional markers to add to the predictive performance of PAPP-A and macrophage inhibitory cytokine-1.
1. Tong S, Marjono B, Brown DA, Mulvey S, Breit SN, Manuelpillai U, et al.. Serum concentrations of macrophage inhibitory cytokine 1 (MIC 1) as a predictor of miscarriage. Lancet 2004;363:129–30.
2. Goddijn M, Leschot NJ. Genetic aspects of miscarriage. Baillieres Best Pract Res Clin Obstet Gynaecol 2000;14:855–65.
3. Redecha P, van Rooijen N, Torry D, Girardi G. Pravastatin prevents miscarriages in mice: role of tissue factor in placental and fetal injury. Blood 2009;113:4101–9.
4. Connolly SJ, Ezekowitz MD, Yusuf S, Eikelboom J, Oldgren J, Parekh A, et al.. Dabigatran versus warfarin in patients with atrial fibrillation. N Engl J Med 2009;361:1139–51.
5. Patel MR, Mahaffey KW, Garg J, Pan G, Singer DE, Hacke W, et al.. Rivaroxaban versus warfarin in nonvalvular atrial fibrillation. N Engl J Med 2011;365:883–91.
6. Muttukrishna S, Swer M, Suri S, Jamil A, Calleja-Agius J, Gangooly S, et al.. Soluble Flt-1 and PlGF: new markers of early pregnancy loss? PLoS One 2011;6:e18041.
7. Duan L, Yan D, Zeng W, Yang X, Wei Q. Predictive power progesterone combined with beta human chorionic gonadotropin measurements in the outcome of threatened miscarriage. Arch Gynecol Obstet 2011;283:431–5.
8. Barrientos G, Fuchs D, Schrocksnadel K, Ruecke M, Garcia MG, Klapp BF, et al.. Low levels of serum asymmetric antibodies as a marker of threatened pregnancy. J Reprod Immunol 2009;79:201–10.
9. Johns J, Muttukrishna S, Lygnos M, Groome N, Jauniaux E. Maternal serum hormone concentrations for prediction of adverse outcome in threatened miscarriage. Reprod Biomed Online 2007;15:413–21.
10. Habayeb OM, Taylor AH, Finney M, Evans MD, Konje JC. Plasma anandamide concentration and pregnancy outcome in women with threatened miscarriage. JAMA 2008;299:1135–6.
11. Tong S, Marjono B, Mulvey S, Wallace EM. Low levels of pregnancy-associated plasma protein-A in asymptomatic women destined for miscarriage. Fertil Steril 2004;82:1468–70.
12. Lam PM, Marczylo TH, Konje JC. Simultaneous measurement of three N-acylethanolamides in human bio-matrices using ultra performance liquid chromatography-tandem mass spectrometry. Anal Bioanal Chem 2010;398:2089–97.
13. Marczylo TH, Lam PM, Nallendran V, Taylor AH, Konje JC. A solid-phase method for the extraction and measurement of anandamide from multiple human biomatrices. Anal Biochem 2009;384:106–13.
14. Coomarasamy A, Truchanowicz EG, Rai R. Does first trimester progesterone prophylaxis increase the live birth rate in women with unexplained recurrent miscarriages? BMJ 2011;342:d1914.
15. Rumbold A, Middleton P, Pan N, Crowther CA. Vitamin supplementation for preventing miscarriage. Cochrane Database Syst Rev 2011;1:CD004073.
16. Wilcox AJ, Weinberg CR, O'Connor JF, Baird DD, Schlatterer JP, Canfield RE, et al.. Incidence of early loss of pregnancy. N Engl J Med 1988;319:189–94.
17. Goetzl L, Krantz D, Simpson JL, Silver RK, Zachary JM, Pergament E, et al.. Pregnancy-associated plasma protein A, free beta-hCG, nuchal translucency, and risk of pregnancy loss. Obstet Gynecol 2004;104:30–6.
18. Taylor AH, Ang C, Bell SC, Konje JC. The role of the endocannabinoid system in gametogenesis, implantation and early pregnancy. Hum Reprod Update 2007;13:501–13.
19. Karasu T, Marczylo TH, Maccarrone M, Konje JC. The role of sex steroid hormones, cytokines and the endocannabinoid system in female fertility. Hum Reprod Update 2011;17:347–61.
20. Wang H, Xie H, Guo Y, Zhang H, Takahashi T, Kingsley PJ, et al.. Fatty acid amide hydrolase deficiency limits early pregnancy events. J Clin Invest 2006;116:2122–31.
21. Maccarrone M, Valensise H, Bari M, Lazzarin N, Romanini C, Finazzi-Agro A. Relation between decreased anandamide hydrolase concentrations in human lymphocytes and miscarriage. Lancet 2000;355:1326–9.
22. Lawrence JB, Oxvig C, Overgaard MT, Sottrup-Jensen L, Gleich GJ, Hays LG, et al.. The insulin-like growth factor (IGF)-dependent IGF binding protein-4 protease secreted by human fibroblasts is pregnancy-associated plasma protein-A. Proc Natl Acad Sci U S A 1999;96:3149–53.
23. Smith GC, Stenhouse EJ, Crossley JA, Aitken DA, Cameron AD, Connor JM. Early pregnancy levels of pregnancy-associated plasma protein a and the risk of intrauterine growth restriction, premature birth, preeclampsia, and stillbirth. J Clin Endocrinol Metab 2002;87:1762–7.
24. Dugoff L, Hobbins JC, Malone FD, Porter TF, Luthy D, Comstock CH, et al.. First-trimester maternal serum PAPP-A and free-beta subunit human chorionic gonadotropin concentrations and nuchal translucency are associated with obstetric complications: a population-based screening study (the FASTER Trial). Am J Obstet Gynecol 2004;191(4):1446–51.
25. Barrett SL, Bower C, Hadlow NC. Use of the combined first-trimester screen result and low PAPP-A to predict risk of adverse fetal outcomes. Prenat Diagn 2008;28:28–35.
26. Marjono AB, Brown DA, Horton KE, Wallace EM, Breit SN, Manuelpillai U. Macrophage inhibitory cytokine-1 in gestational tissues and maternal serum in normal and pre-eclamptic pregnancy. Placenta 2003;24:100–6.
27. Segerer SE, Rieger L, Kapp M, Dombrowski Y, Muller N, Dietl J, et al.. MIC-1 (a multifunctional modulator of dendritic cell phenotype and function) is produced by decidual stromal cells and trophoblasts. Hum Reprod 2012;27:200–9.