Foetal loss after chorionic villus sampling and amniocentesis in twin pregnancies: A multicentre retrospective cohort study

Abstract Objective We aimed to determine foetal losses for DCDA and MCDA twins following transabdominal CVS or amniocentesis performed <22+0 weeks. Methods Retrospective cohort study conducted in the UK and Belgium 01/01/00–01/06/20. Cases with unknown chorionicity, monochorionic complications or complex procedures were excluded. Uncomplicated DCDA and MCDA twins without invasive procedures were identified as controls. We reported foetal losses <24+0 weeks and losses of genetically and structurally normal foetuses. Results Outcomes were compared for DCDA foetuses; 258 after CVS with 3406 controls, 406 after amniocentesis with 3390 controls plus MCDA foetuses, 98 after CVS with 1124 controls, and 160 after amniocentesis with 1122 controls. There were more losses <24+0 weeks with both procedures in DCDA (CVS RR 5.54 95% CI 3.38–9.08, amniocentesis RR 2.36 95% CI 1.22–4.56) and MCDA twins (CVS RR 5.14 95% CI 2.51–10.54, amniocentesis RR 7.01 95% CI 3.86–12.74). Losses of normal foetuses were comparable to controls (DCDA CVS RR 0.39 95% CI 0.05–2.83, DCDA amniocentesis RR 1.16 95% CI 0.42–3.22, MCDA CVS RR 2.3 95% CI 0.71–7.56, and MCDA amniocentesis RR 1.93 95% CI 0.59–6.38). Conclusions This study indicates increased foetal losses for DCDA and MCDA twins following CVS and amniocentesis with uncertain risk to normal foetuses.

� Procedures themselves may not contribute significantly to pregnancy loss for women at a higher baseline risk but may be more influential for women at lower a priori risk.

What does this study add?
� Spontaneous losses of twins prior to 24 weeks are increased after CVS and amniocentesis for both DCDA and MCDA pregnancies compared to controls.
� Losses of structurally and genetically normal twins appear comparable to control groups.
However, due to low numbers of adverse outcomes in our cohort and other published analyses, further evidence is required.
� Maternal counselling should highlight different baseline risks between MCDA and DCDA twins and the contribution of maternal and pregnancy factors to outcomes for complex twin pregnancies.

| INTRODUCTION
Despite advances in non-invasive testing, chorionic villus sampling (CVS) and amniocentesis are the prenatal tests of choice for diagnosing chromosomal and genetic abnormalities. Assessment in a randomised controlled trial (RCT) is the gold standard to estimate a procedure-related risk. However, it is unlikely that appropriately powered, high-quality randomised controlled trials, evaluating comparative effects of CVS and amniocentesis would be feasible or acceptable to patient groups.
A recent meta-analysis reported a weighted pooled procedurerelated risk of miscarriage for singletons of 0.35% for both CVS and amniocentesis. 1 Accepting the limitations of observational data and relative imprecision of the estimated risks, recent RCOG guidance concluded that procedure-related risk of miscarriage for singleton pregnancy is likely to be less than 0.5%. 2,3 Evidence quantifying procedure-related risks in multiple pregnancies is significantly more limited. Observational studies published in the last 2 decades suggest that the procedure-related risk of miscarriage in twins may be lower than previous estimates. 4 A systematic review published in 2020 included 2713 amniocentesis procedures and 349 CVS in multiple pregnancies. The results indicated no significant difference in foetal loss before 24 weeks gestation and within 4 weeks of procedure, compared to those not undergoing an invasive procedure. These findings are further supported by a recent large multicentre retrospective cohort study that used logistic regression.
The authors found no significant contribution of CVS to the risk of postprocedure miscarriage. Interestingly, when the same dataset was analysed using propensity scoring, there was a 3.5% higher absolute risk of foetal loss following CVS for women at a low baseline risk. 5,6 Therefore, more evidence is needed to establish a robust evidence base and clarify inconsistent study findings.
Attempts to quantify procedure-related risks in multiple pregnancies are challenging. Published case series tend to be small with variation in definitions of procedure-related foetal loss. 4 Monochorionicity, with shared vascular connections, confers additional complexity. 7 Increased baseline risks, technical challenges, and postprocedure losses may be attributable to the monochorionicity rather than the procedure itself and yet this information is often unavailable or inadequately reported.
To improve the clinical utility of the available data, we decided to analyse dichorionic diamniotic (DCDA) and monochorionic diamniotic (MCDA) twin pregnancies as two distinct cohorts from the outset. For each group, we sought to identify structural and genetically normal foetal losses and look for phenotypic differences in types of foetal loss.
TTTS, or discordant structural or genetic abnormalities were excluded from control groups. Pregnancies affected by foetal loss between 11+ 0 and 15+ 0 weeks were excluded from the amniocentesis control group.
Women who had undergone CVS (11+ 0 -22+ 0 weeks) or amniocentesis (15+ 0 -22+ 0 weeks) in six centres between 01/01/00 and 01/06/20 were considered as the study group. Exclusion criteria included pregnancies where chorionicity could not be determined or with unobtainable outcomes. MCMA pregnancies, those with TRAP, and TTTS were excluded. We also excluded pregnancies when more than two attempts were required to obtain a sample, when procedures were performed alongside embryo reduction or when selective foetocide was performed within the reporting period (<24+ 0 weeks).
Twin chorionicity and amnionicity were determined by ultrasound scan in the first trimester as per evidence-based guidance. 9,10 CVS and amniocentesis were carried out by foetal medicine specialists, using continuous ultrasound guidance and aseptic technique.
17-21G needles were used for CVS, 20-22G for amniocentesis. All CVS procedures were performed transabdominally and intraamniotic dye infusion was not used for amniocentesis. Foetal wellbeing was assessed upon procedure completion.
For controls, CVS, and amniocentesis groups, we chose to assess losses at 24 +0 weeks as, below this threshold, twin foetuses are unlikely to be viable. For controls, CVS, and amniocentesis groups, we reported overall foetal losses <24+ 0 weeks, including deaths of both foetuses and single foetal deaths and any reported deaths of structurally and genetically normal foetuses <24+ 0 weeks. For CVS and amniocentesis groups, we also reported overall foetal losses <24 + 0 weeks within 15 days of procedure, including death of both foetuses and single foetal deaths. We also reported any deaths of structurally and genetically normal foetuses within 2 weeks (<15 days) of the procedure as more likely to be related to the procedure. Forest plots to compare our data to those recently published were generated in Revman 5.4 (The Cochrane Collaboration, available at revman.cochrane.org). Risk ratios with random effects were calculated for CVS and amniocentesis in DCDA and MCDA twins, respectively. Where judged appropriate, data were pooled. 2

| RESULTS
Changes in patient record systems in BWH and UHB meant that data for women who had invasive procedures could only be obtained between 01/01/06 and 30/09/19 in BWH and between 01/01/09 and 01/06/20 in UHB. In total, 899 women with twin pregnancies had CVS or amniocentesis during the study period. After exclusions, 129 women with DCDA twins had CVS and 203 had an amniocentesis.
After exclusions, 49 women with MCDA twins had CVS and 80 had amniocentesis ( Figure 1).
Eleven foetuses died within 2 weeks of CVS, including two twin pairs, and seven foetuses survived by their co-twin (Table S1a).

-
There were no losses of structurally and genetically normal foetuses in the CVS group, which were comparable to losses that occurred in controls (1.5%, 50/3406, RR 0.39 95% CI 0.05-2.83) ( Three foetuses died within 2 weeks of amniocentesis, including one twin pair and one foetus survived by their co-twin (Table S1b).
Overall, there were two losses of pairs of genetically and structurally normal twins, including the twin pair that died within 2 weeks of amniocentesis (

| MONOCHORIONIC DIAMNIOTIC TWINS (MCDA)
In Seven foetuses demised within 2 weeks of CVS including, three twin pairs and one foetus that was survived by their co-twin (Table S2a).
Overall, there were three losses of structurally and genetically normal foetuses, none of which occurred within 2 weeks of CVS (
Two cases required a second attempt to obtain the sample. Fifteen foetuses died within 2 weeks of the procedure, one that was survived by their co-twin and seven twin pairs (Table S2b).
There were three losses of genetically and structurally normal foetuses alongside an affected co-twin, two of which occurred within 2 weeks of amniocentesis (Table S2b).

| Main findings
This retrospective cohort study demonstrates that spontaneous losses of twins prior to 24 weeks are increased after CVS and amniocentesis for both DCDA and MCDA pregnancies compared to controls. Whilst losses of structurally and genetically normal twins appear comparable to control groups, the confidence intervals are very wide, and therefore, the data must be interpreted cautiously.
Our data are compatible with no difference in risk of normal twin demise between cases and controls, but also with clinically significant differences in both directions. Whilst a clinically important increase in risk cannot be ruled out, there is a possibility that the risk is actually decreased for those undergoing prenatal diagnosis.

| Strengths and limitations
Our multicentre retrospective cohort study includes a relatively large number of DCDA and MCDA twin pregnancies and reflects the provision of contemporary specialised prenatal diagnosis for complicated twin pregnancies with advances in ultrasound technology and invasive techniques. Strict exclusion criteria were applied to ensure that twin subpopulations with higher intrinsic risks of foetal loss did not impact overall outcomes, including MCMA pregnancies, and those affected by TRAP and TTTS. Importantly, we have reported outcomes by chorionicity, described procedural factors, and provided detailed relevant foetal information for all post-procedure demises (Supplemental Tables S1 and S2).
An important limitation of our dataset is considerable demographic data loss due to under-reporting of maternal BMI, ethnicity, and parity, which is a limitation in describing a priori risk of loss. This was largely due to the inclusion of tertiary referral cases and changes in patient information systems during the study period.
Despite efforts made to follow-up outcome data from referring units, some data were not recoverable. It is possible that further adverse outcomes were contained within these missing data.

| Interpretation
The best evidence regarding procedure-related risks for CVS and amniocentesis will continue to come from observational data. The accumulated evidence in singleton pregnancies indicates that CVS and amniocentesis pose a reassuringly small additional risk of pregnancy loss. 13 The limiting factors for such estimates in twin pregnancies are relatively low numbers of procedures and very low event rates. Data are especially limited for monochorionic twin pregnancies. A recent systematic review of singletons included 13 times the number of procedures that were included in a similar systematic review in twins. 13,14 Monozygosity and monochorionicity confer additional risks that mark these twin pregnancies as a distinct population. 7 Our data indicate a similar risk of post-procedure demise for CVS and amniocentesis in MCDA twins; however, these risks are appreciably higher than those NAVARATNAM ET AL.
we observed in DCDA twins (Tables 2a and b). This observation supports our assertion that dichorionic or pooled twin outcomes should not be extrapolated to the monochorionic population for the purposes of evidence-based guidance and patient counselling.
Since the most recent systematic review of procedure-related loss in twin pregnancies, two further observational datasets have been published. 5 losses prior to 24 weeks gestation increased two-fold following CVS, but the procedure itself did not contribute significantly to the risk of foetal loss. 5 The authors attributed the increase in foetal losses to maternal and pregnancy factors. 5 When, in an attempt to deal with any potential confounding of maternal and pregnancy factors, the same dataset was analysed using propensity score matching, Gil et al.
reported a significant (3.5%) increase in the individual risk following CVS for women at the low baseline risk.

| Clinical and research implications
Our findings of increased post-procedure losses following CVS and amniocentesis in both DCDA and MCDA twin pregnancies and possible clinically important risks to normal twins stand in contrast to the conclusions of recent published studies.
It is important to stress that data derived from small studies with significant heterogeneity will always limit precision of summary statistics. Even when data are sufficiently homogeneous for clinically meaningful pooled analysis, lack of statistical significance does not exclude a possibility of clinically important differences. This is particularly relevant when dealing with low event rates causing bias towards statistical non-significance. When the 95% confidence interval includes 1, the 'logical' conclusion is that there is no difference in procedure-related loss. It should be noted, however, that our data are compatible not only with an increase in procedure-related loss but also with the possibility of a clinically significant 'protective' effect from invasive procedures. Whilst this could be ignored as implausible, it is not inconceivable that women with MCDA twins undergoing prenatal diagnosis may relax after testing and rest from work. They may benefit from enhanced care with expert imaging and increased surveillance. We encourage future authors not to focus solely on the statistical significance, but to interpret the data much more holistically.

| CONCLUSION
This study indicates increased twin losses prior to 24 weeks following CVS and amniocentesis in MCDA and DCDA twin pregnancies and these findings contrast somewhat with recently published evidence.
The uncertain procedure-related risk to structurally and genetically normal twins persists due to low numbers of adverse outcomes in our cohort and other published analyses. Any counselling should highlight different baseline risks between MCDA and DCDA twins and the contribution of maternal and pregnancy factors to outcomes for complex twin pregnancies.