MULTICENTRE RANDOMISED CLINICAL TRIAL OF CHORION VILLUS SAMPLING AND AMNIOCENTESIS

MULTICENTRE RANDOMISED CLINICAL TRIAL OF CHORION VILLUS SAMPLING AND AMNIOCENTESIS First Report

CANADIAN COLLABORATIVE CVS-AMNIOCENTESIS CLINICAL TRIAL GROUP*

Summary

who were aged 35 years or expected date of delivery were chorionic villus sampling (CVS) at 9-12 2787

women

more at

randomised to weeks gestation or amniocentesis at 15-17 weeks for the detection of a chromosomal abnormality in the fetus. 396 women were excluded after randomisation because of a non-viable fetus, a multiple pregnancy, or infection or because the pregnancy was too far advanced (more than 12 completed weeks). Among all women eligible at the time of the first study procedure there were 89/1169 (7·6%, 95% confidence interval [CI] 6·2-9·3%) total losses

*The members of the clinical trial group were: D. WILSON,† B. MCGILLIVRAY, D. KALOUSEK, C. GANSHORN, J. HALLt (Vancouver); P. OLLEY, the late P. BOWEN, I. TATARYN, M. CHOI, D. REID, Z. PYATT (Edmonton); N. RUDD,† G. CARSON, D. COX, M. L. KLIMEK (Calgary); J. EVANS, J. HAHERTON,† M. RAY, K. MACDONALD, C. HARMON, J. JOHNSON (Winnipeg); R. NATALE, F. SERGOVICH, J. SUTHERLAND (London); P. MOHIDE, P. FULLER,† D. TOMKINS,† I. UCHIDA, V. HUNTER, D. FEENY (Hamilton); J. SHIME,† E. HUTTON,† I. TESHIMA, V. MARKOVIC,

J. BROWN, T. A. DORAN (Toronto); A. HUNTER,† H. MUGGAH, R. BENZIE,† H. WANG, D. THOMPSON, J. GORDON (Ottawa); L. DALLAIRE,† M. DESROCHERS, M. LEMAY (Montreal, University of Montreal); A. LIPPMAN,† M. VEKEMANS,† L. CARTIER, T. PERRY, L. PINSKY, P. FOURNIER, E. HAMILTON (Montreal, McGill University); R. GAGNE, M. CHATILLON

(Laval).

The data centre was staffed by: R. MILNER† M. JOHNSTON, and M. TAYLOR Members of the trial executive and

Hamerton) are indicated.†

implementation

group

(chairman, J.

(spontaneous and induced abortions and late losses) in the CVS group and 82/1174 (7·0%, CI 5·6-8·6%) in the amniocentesis group for an excess of 0·6% for those women undergoing CVS, with a tendency to later losses in the CVS group. Approximate 95% CIs indicate that this difference is most unlikely to be greater than 2·7%. Mean birthweights for each week of gestation were similar in both groups, with no evidence of excess small-for-gestational age babies in the CVS group. The proportion of preterm births were similar in both groups. Perinatal mortality was greater in the CVS group, the greatest imbalance being beyond 28 weeks. Preliminary analysis has not disclosed an obvious recurrent event in the CVS group which might explain a cause-andeffect relationship for these late fetal losses. There is no evidence of any excess intrauterine growth retardation babies in the CVS group. Maternal morbidity was similar in both groups. 103/1037 (9·9%) women eligible for CVS had a further amniocentesis; 32 of these second procedures were needed to complete the cytogenetic diagnosis. More problems such as confined mosaicism and maternal cell contamination occurred in the interpretation of the CVS samples, but these were clarified by amniocentesis when appropriate. Introduction TRANSCERVICAL chorionic villus sampling (CVS) was introduced over fifteen years ago as a potential method for the first-trimester prenatal diagnosis of selected genetic disorders.’ Developments in ultrasound have allowed CVS to be made widely available, and more than 50 000 clinical diagnostic procedures have been reported from over a hundred centres worldwide.22 Although estimates of the complication and fetal loss rates associated with CVS soon proliferated none was derived from a randomised controlled trial. To ensure that CVS was not introduced in Canada before its risks and safety had been assessed, a multicentre randomised trial was started in 1984.

2

Eleven centres took part in a trial designed to compare fetal loss rates (including induced abortion) in women having amniocentesis and women having CVS. The accuracy of both procedures was also examined. This is the first report of results from this study. 2979 women, who were unable to obtain CVS in Canada outside the trial, were assigned randomly to one or other procedure. The results presented here are limited to those for the 2787 women in whom the indication for prenatal diagnosis was an increased risk of having a child with a chromosomal abnormality due to late maternal age. Patients and Methods Patients Women who would be 35 years of age or older at the expected date of delivery (and women with other indications for prenatal diagnosis) were counselled about the randomised study. To be eligible a woman had to be at less than 12 completed weeks’ gestation and able to attend for ultrasound scans at 9-12 weeks and at 15-17 weeks. Women were excluded if there was a fetal loss before the first ultrasound or if a dead or disorganised fetus was detected at the first scan, if there was a multiple pregnancy, if they were Rh isoimmunised, or if they had an untreated cervico-vaginal infection. Women were also ineligible if the first scan showed that the gestation was 12 weeks or greater. Ineligibility was assessed by the collaborating centres, though a few women were declared ineligible by the data centre on the basis of the information available and the criteria developed for the trial. The study design was approved by all the ethics committees and all women enrolled provided verbal and written consent.

Study Design In preparation for the trial, standards were set up to ensure that obstetricians and cytogenetics laboratories had reasonable experience. Obstetricians had to do at least 30 procedures and be successful in obtaining 10 mg or more of chorionic villi in 23 out of 25 consecutive cases. Cytogenetics laboratories had to obtain a report based on 15 banded cells in 19 out of 20 consecutive adequate

samples. Randomisation was done centrally after stratification by centre and by maternal age (35-38 and 39 or more). All participants were scheduled by centre coordinators to have their first allocated procedure(s) at between 9 and 12 weeks. For women assigned to CVS, the 9-12 week procedures were an ultrasound scan (to confirm fetal age and viability) and CVS. For women assigned to amniocentesis the 9-12 week procedure was the ultrasound scan alone. At 15-17 weeks all women had a second scan, to detect fetal neural tube defects; the CVS group had blood taken for a maternal serum x-fetoprotein (AFP) measurement while those allocated to amniocentesis had this procedure done. Participants were asked to report all spontaneous and induced abortions to the coordinator at their centre. Coordinators also telephoned all participants still in the study at 21-23 weeks gestation to obtain information about complications or problems arising since the most recent procedure. When a miscarriage was reported a form was completed. A 42-46 week follow-up was done to obtain information about delivery. A card was sent to all the remaining women about 3 weeks before the due date. They were asked to take the cards with them to hospital for physicians to complete. Only very basic information about delivery, gestational age, birthweight, and any abnormalities obvious at birth was sought.

Obstetric

Techniques

CVS was done, according to the usual practice at each centre, by ultrasound guided transcervical aspiration. Seven centres used Portex catheters, three ’Vascath’ catheters, and one centre used both. Cervical cultures were obtained before introduction of the

centre to centre

in respect of aspirating, syringe

size, and amount of

pressure. No more than three aspiration attempts were allowed to obtain 10 mg or more of chorionic tissue. A new catheter was used for each aspiration. The patients were then instructed to have reduced activity for 24 hours after the procedure. If CVS was unsuccessful, the decision to try again was made by the obstetrician and the woman. If CVS was decided against, amniocentesis was

negative

offered. The size of amniocentesis needles varied. Three

centres

using

20G, four using 22G, one used both 21 and 22G, and another two used both 20 and 22G. One centre where CVS was done did not do amniocenteses: these were done at other hospitals and needle sizes were not reported. The use of ultrasound during amniocentesis varied. The volume of amniotic fluid removed varied from 5 to 40 ml. These patients also were advised to reduce activity for 24 hours. If the pregnancy was terminated after prenatal diagnosis of a chromosomal abnormality with either procedure, laboratory confirmation was requested. If a woman continued with her pregnancy, the baby was examined for confirmation of the prenatal

diagnosis.

Cytogenetic Techniques The cytogenetics laboratories chose their own preferred methods for the preparation of chorionic villus and amniotic fluid samples. All thoroughly dissected chorionic villus samples to remove contaminating decidual cells. Most laboratories received amniotic fluid in two equal amounts to permit, when necessary, the identification of maternal cell contamination of the initial aspirate. For cytogenetic analysis of CVSmaterial both cytotrophoblastic preparations (made directly from villi without incubation, with overnight incubation in the presence of colchicine, or with incubation for less than a week of villi treated with colchicine at the end of incubation) and cultured preparations (made from villi cultured for more than a week, usually after enzymatic treatment to remove the syncytiotrophoblastic and cytotrophoblastic cell layers, leaving the mesenchymal core for proliferation in culture). For cytogenetic analysis of amniotic fluid samples preparations were made by in situ harvesting of colonies of amniotic fluid cells without any subculturing and by suspension harvesting or in situ harvesting after subculturing. All laboratories used their routine banding techniques for karyotypic analysis and any other techniques, as required, for interpretation of cytogenetic findings. The number of cells analysed completely was at the discretion of the laboratory, although it was recommended that at least 15 cells be analysed for all cytogenetic diagnoses. Problems with the cytogenetic diagnosis, such as inadequate material, inadequate banding, mosaicism, and maternal cell contamination, were recorded. The laboratory report included the final diagnosis, which was 46,XY, 46,XX, abnormality (with karyotype), or failure/not done. Detailed cytogenetic results were reviewed for all reported abnormalities and problems of mosaicism and maternal cell contamination.

Statistical Methods The major focus of this trial was the total fetal loss rate in women for whom CVS was a realistic option. Consequently, in the initial analysis, all subsequent losses in women who had a viable fetus at the time of the first procedure, whether the loss resulted from termination of pregnancy or miscarriage, were counted as "outcome" events. Rates of loss were calculated with exact 95% confidence intervals (CI) whether or not the women completed all aspects of the protocol. The power of this study to detect the hypothesised difference of 3% in overall loss rates between groups was also calculated and the approximate CI for the difference in observed total loss rates between groups was calculated. To ensure that post-randomisation exclusions of ineligible women did not introduce bias, the difference between groups in total rates of fetal loss was calculated for all women randomised whether or not they remained in the study after randomisation, whether or not they actually received the allocated procedure, and whether or not they provided full follow-up data.

3

Results

TABLE I-COMPARABILITY OF GROUPS (ELIGIBLE PATIENTS)

2787 women were randomised, 1391 to CVS and 1396 to amniocentesis (figure). Because of differences in clinical practice between centres it was not always possible to do an ultrasound scan to assess a woman’s eligibility before randomisation, and subsequent to randomisation 396 women were declared ineligible because of a non-viable fetus, multiple pregnancy, or gestation advanced beyond 12 completed weeks. These patients were excluded from the analysis of outcomes among eligible women, but were included in an analysis of outcomes among all randomised women. The total number of ineligible women was much the same in the two groups (figure). A further 411 women did not complete the full protocol, for a variety of reasons, but they were counted in all analyses based on all eligible women.

The two groups were comparable (table I). The average age in both groups was 38 and over 96% of women were having the procedure for maternal age reasons only. The frequency of prior miscarriage or therapeutic abortion was similar in the two groups. Medical problems such as hypertension and diabetes were equally rare in both groups. OBSTETRIC FINDINGS

In the amniocentesis group the time between randomisation and the first scan was, on average, 2days shorter than it was in the CVS group (table 11). Mean gestational age at first procedure was 75-0 for the CVS group and 73-7 days for the amniocentesis group, on the basis of last menstrual period (LMP). Gestational age estimated by ultrasound was consistent with menstrual dating in 88% of cases.

Placental location was equally divided between anterior and posterior, 15 % in each group being in another location. At the first ultrasound 75 dead or disorganised fetuses were found in the CVS group and 57 in the amniocentesis group. All 132 were declared ineligible and contribute only to the analysis of all randomised women. Of the 1037 women undergoing CVS, 103 (9-9%) also had an amniocentesis because of difficulties with the CVS procedure itself or with the laboratory analysis. Of the 245 women assigned to but not receiving amniocentesis 31 had a miscarriage before the procedure, 6 had their pregnancies terminated, and 208, though eligible,

Randomisation and eligibility.

did not complete the protocol after randomisation. Women assigned to amniocentesis refused further participation more often than women assigned to CVS (13% vs 9%).

the time had the second (table III). In both 1 of scans indicated a non-viable fetus. fewer than % groups Information was available at 15-17 weeks on complications since the first procedure for 70% of cases seen at 9-12 weeks. As expected, there was an excess of bleeding and spotting reported by the CVS group. Reports of other maternal morbidity, such as fever, were similar between the groups, but this information was not elicited rigorously. Because of initial concern about loss rates among women able to have either of the allocated procedures, attention focused first on outcomes among eligible women with a viable fetus at 9-12 weeks. There were 89 (7-6%) losses (miscarriages, terminations, and stillbirths) among women allocated to CVS and 82 (7-0%) losses among women allocated to amniocentesis (table IVA). 95% CIs are 6-2-9-3% and 5.6-8.6 %, respectively. The difference in loss rate is not significant. The timing of losses in the two groups is generally similar. However, there appears to be a shift to

90% of women still pregnant

study ultrasound

scan at

at

15-17 weeks

TABLE II-NON-OUTCOME DETAILS OF FIRST

PROCEDURE(S)

*History obtained for fewer than 70%

cases.

(9-12 WEEK)

4 TABLE IV-OUTCOME

I

*Excludes

women

found to be

(ENDPOINTS)

IN

(A) ELIGIBLE AND (B) ALL RANDOMISED WOMEN

i

I

ineligible post randomisation at or before first study procedure.

later losses in the CVS group which, although not significant, merits further attention. There was 1 neonatal death in each group. Combining late losses (beyond 20 weeks) and neonatal deaths yielded crude perinatal mortality rates of 1-3% in the CVS group and 0-6% in the amniocentesis group. The greatest imbalance in the frequency of stillbirths was after 28 weeks, when there were 6 in the CVS group and only 1 in the amniocentesis group. No event in common that might explain these late losses has so far emerged. Because of concern that the post-randomisation exclusions of ineligible women might have introduced bias, outcomes have also been analysed for all randomised women, irrespective of subsequent eligibility, the only exclusions here being 53 women completely lost to followup. There were 230 (16-8%) induced and spontaneous losses in the 1368 women allocated to CVS and 206 (15 -1%) losses in the 1366 women allocated to amniocentesis (table IVB). The higher CVS loss rate is not significantly different from the amniocentesis rate and the confidence limits overlap (14-2-18-2% and 13.2-17-0%). Thus when loss rates among all randomised women were considered, the results were similar to those seen in the analyses of eligible

for whom

follow-up data

still missing. In the CVS group, by contrast, women who did not complete the study were outnumbered by women for whom data are incomplete. The potential for bias created by these inequalities underscores the need to include all eligible women in the analyses. are

CYTOGENETIC FINDINGS

The cytogenetic results for 947 CVS samples and 933 amniotic fluid samples received for analysis are shown in tables v-ix. On average, 13 and 23 days were required for completion of the cytogenetic analysis of chorionic villi and amniotic fluids, respectively. Laboratory failure was significantly more frequent for CVS (table v; 1 5 % vs 0.1 % for amniocentesis, p < 0-001). TABLE VI-CYTOGENETIC ABNORMALITIES FOUND ON ALLOCATED

PRENATAL PROCEDURE

only. Birthweights are available for 870/1191 (73-0%) of CVS and 810/1200 (67-5%) of amniocentesis cases. There appears to be no difference in birthweight distribution or mean birthweight (CVS 3466 g; amniocentesis 3451 g). There is no increase in intrauterine growth retardation in the CVS group and the proportion of preterm deliveries is the women

in both groups. The outcome of pregnancy is known for 93% of the total study population. Information is still being sought and the final proportion will be higher. Although outcomes are unknown for similar proportions of women assigned to each procedure (table IV), these proportions comprise different categories of women. In the amniocentesis group, half were women who did not complete the study and half are women same

*Only indication for prenatal diagnosis

was late maternal age, except for 1 CVS case where there was also a familial translocation. All balanced rearrangements diagnosed prenatally were familial, except for 1 amniocentesis case where 2 de novo translocations were diagnosed. tl was diagnosed on amniocentesis after obstetric failure to obtain chorionic villus sample. 1 diagnosed on amniocentesis after a laboratory failure to obtain a diagnosis on CVS (only three cells 47,XY, + E). t2 mosaics. §All cases of mosaicism where only one abnormal cell was observed were excluded from both CVS and amniocentesis abnormal diagnoses.

TABLE VII-ABNORMALITIES CONFINED TO THE CHORIONIC VILLI TABLE V-CYTOGENETIC DIAGNOSIS ON ALLOCATED PRENATAL

OR AMNIOTIC FLUID CELL CULTURES

(FALSE POSITIVES)

PROCEDURE

i i *See table VI for types of cytogenetic abnornality. t5 laboratory failures subsequently found to be 46,XX and 8 46,XY. 1 case o laboratory failure was later found to have an abnormality on amniocentesis.

*Numbers leading to follow-up amniocentesis are m parentheses. tl level III mosaicism: (46,XY/47,XY,+20). 2 were level II: 46,X,del[X] (q22-qter) and (46,XX/4b,XX,t[2;12] (q33;p13),

(46,XX/

5 TABLE VIII-MATERNAL CELL CONTAMINATION FOUND ON ALLOCATED PRENATAL PROCEDURE

I

I

*All but 1 problem with maternal cell contamination in chorionic villus samples occurred with CVScultured preparations. TABLE IX-CHORIONIC VILLUS SAMPLING LEADING TO FOLLW-UP AMNIOCENTESIS

_________________________ abnormality resulted in spontaneous abortion before follow-up amniocentesis : (47,XY, + 16). *1 other pregnancy with confined

Because

CVS

and

amniocentesis

are

alternative

procedures for obtaining samples for prenatal diagnosis diagnostic accuracy is paramount. There was no difference between CVS and amniocentesis with respect to the proportion of males and females with normal karyotypes (0-537 and 0463 for CVS, and 0-536 and 0-464 for amniocentesis, respectively). The percentage of abnormal cytogenetic diagnoses, however, was significantly higher for CVS (4-6% vs 2-4% for amniocentesis; p < 0-01). The difference in abnormality rate between the two prenatal procedures can be explained, for the most part, by the significantly higher proportion of "false positive" diagnoses for CVS (tables VI and VII; 2-0% for CVS vs 0-3% for amniocentesis; p < 0-001). False positives are defined as those cases where an abnormality was detected on prenatal diagnosis that was not found in the follow-up amniocentesis, abortus, or baby. With 4 exceptions (1 mosaic 46,XX/ 47,XX, + mar; 2 mosaic for a polyploid line; and 1 47,XX, + 16 with spontaneous abortion before retesting), all CVS false-positive cases led to amniocentesis, where the correct diagnosis was made. In the amniocentesis material, the 3 false positives were all correctly interpreted. There was also a significantly higher frequency of true positive aneuploidy for CVS than for amniocentesis (2-1% vs 1 -0%, p < 0-05). This difference may be explained by the earlier CVS diagnosis of aneuploid fetuses that would otherwise be aborted spontaneously before amniocentesis.3 The number of unbalanced and balanced structural rearrangements (eg, translocations, deletions, duplications, inversions, and isochromosomes) was too small to reveal any difference between CVS and amniocentesis. Discussion Two issues in the design and analysis of this (or any other) study of the risks of CVS must be borne in mind when evaluating the results. Because terminations following the diagnosis of a cytogenetic abnormality might remove some cases from the CVS group which were destined to abort spontaneously, a comparison restricted to spontaneous losses after 9-12 weeks in both groups would be misleading. Furthermore, because a pregnancy might be terminated for obstetric reasons related to the diagnostic procedures (eg, uterine infection) it cannot be assumed that it is only

spontaneous losses that constitute the "cost" of the prenatal diagnostic method used. All losses post procedure must be counted. Second, because of the uncontrolled (and uncontrollable) occurrence of spontaneous and induced abortion in women planning to have and actually having prenatal diagnosis, an absolute rate of loss due to CVS cannot be calculated. To do this would require a comparison of women randomly assigned to CVS or to no intervention-with either no-one allowed to have an induced abortion for any reason or the complete karyotyping of all spontaneous fetal losses. The former is unethical; the second, impossible. Fortunately, clinically useful information on the difference in risks between procedures can be reliably estimated with the study design of the Canadian national trial. Among all women randomised the difference observed between groups was 1-7% with an upper confidence limit of 40%. The observed difference in loss rates among women who were eligible and had a viable fetus at 9-12 weeks was 0-6%. In neither case was the difference between CVS and amniocentesis significant statistically. If the true difference in losses between the eligible groups was 3% or greater, there would have been an 84% probability of detection, given the size of the study. Moreover, from this analysis of data for women aged 35 or more at delivery, who had a viable fetus at less than 12 weeks gestation, and who were eligible for CVS, we can be confident that the real difference in loss rate is between - 1,5% and + 2-7% (the CI around the observed 0-6% difference between groups). The results suggest that, under conditions similar to those pertaining during this trial, a woman of 35 or older with a viable fetus at the time of CVS is unlikely to have a total loss rate of more than 9-3% (the upper 95% CI on the observed rate in eligible women allocated to CVS). The figure for amniocentesis is unlikely to be more than 8-6%. Although we cannot be certain by how much CVS increases fetal loss rates above some natural background we can be confident that the difference in loss rates between CVS and amniocentesis is not statistically significant. However, the increase in crude perinatal mortality rate after CVS, while not significant, is disturbing. In

deciding

to

undergo prenatal diagnosis

women

than just fetal loss rates. Factors such as the likely success of the procedure on the first attempt and the accuracy of the results are also likely to influence their choices. The technical limitations on cytogenetic analysis were inadequate numbers of cells and/or inadequate banding. These problems were more frequent with CVS (8% and 10%, respectively) than with amniocentesis (2% and 3%, respectively), even though a constant effort has been made at improving cytogenetic methods with CVS.’ Two major problems in prenatal diagnosis that require careful cytogenetic interpretation are confined mosaicism (table vn) and maternal cell contamination (table VIII). Confined mosaicism (pseudomosaicism) occurs in amniotic fluid cell cultures and may be interpreted as a false positive when only one cell is found with the abnormal karyotype (level I mosaicism) or when the cells from one flask or clone are found with the same abnormality (level II). Level I and level II moscaicism are seldom confirmed in the fetus.s When there are multiple clones or cells from more than one flask with the same abnormality (level III) this may or may not represent fetal mosaicism. In our study, there was only 1 case of level III mosaicism on amniocentesis and it was correctly interpreted as a false positive.

consider

more

6

In studying chorionic villi confined mosaicism may arise during rapid cell proliferation in the early embryo and result in a cell line which occurs in only some tissues of the conceptus.6 If only extraembryonic tissues are involved this could lead to a false-positive result; if restricted to the fetus a false-negative result may be reported. The cells of the cytotrophoblast are derived from the trophoblast at the morula stage and are highly mitotically active.’ The cells used for cultured preparations are derived from the inner cell mass, specifically from the three or four cells that give rise to the embryo proper, the mesenchymal core of the chorionic villi.8 Thus discrepancies between karyotypes in cells of trophoblast, mesenchyme, and fetus are possible. Our results are consistent with published observations that mosaicism can be confined to cytotrophoblastic preparations, mesenchymal preparations, or both cytotrophoblastic and mesenchymal preparations, without appearing in the fetus.99

Maternal cell contamination in amniotic fluid cell cultures is usual but is more likely when the first few millilitres of fluid are used for tissue culture.10 The 4 instances in this study were all interpreted correctly because of the presence of male cells or, in 1 case, abnormal cells. Maternal cell contamination is more common in CVS preparations, especially in culture." Maternal cells are intimately associated with the villi and proliferate rapidly in vitro. They can become the only cells growing in a culture preparation, so the finding of 46,XX cells in cultured villi alone may not represent the fetal karyotype. Confirmation in another CVS preparation which has not been cultured or, where necessary, in a follow-up amniocentesis may be required. In this study there were 19 cultured CVS preparations and 1 cytotrophoblastic preparation with recognised maternal cell contamination where cells with a male karyotype were found in the same or other preparations from the sample (20/478=4-2%). However, in 4 cases, maternal cell contamination was suspected because there were only female results from the culture preparation and there was a possibility that the sample contained maternal decidua. Amniocentesis resulted in 2 female karyotypes and 2 male karyotypes, confirming the suspicion of maternal cell contamination. Because the analysis of outcome has not yet been completed we do not know how many cases there were where maternal cell contamination was not suspected but there was a male or abnormal liveborn. Thus there were at least 2 cases where the CVS cytogenetic diagnosis was made incorrectly on maternal cells. With thorough dissection of villi and enzymatic pretreatrnent11,12 it may be possible to reduce the frequency of maternal cell contamination in cultured CVS. Although there were more problems associated with the interpretation of cytogenetic findings for CVS than for amniocentesis, it was possible to reach a definitive diagnosis in all cases by offering follow-up amniocentesis. About 10% of the women undergoing CVS (103 cases) had a follow-up amniocentesis (table IX): 64-1% resulted from a failure to obtain a chorionic villus sample and 32% from cytogenetic problems (there were 4 for other reasons). Thus if CVS is to be routinely offered, it may need to be supplemented by amniocentesis in some cases, though the rate of follow-up amniocentesis may fall with experience in CVS and cytogenetic interpretation of mosaicism.13 The interpretation of CVS findings currently imposes on cytogeneticists, clinical genetic staff, and patients a bigger burden than amniocentesis does. Cytogenetic findings in different types of CVS preparation must continue to be

recorded, with confirmation of abnormalities in the outcome, amniocentesis. pregnancy

to

reduce

need

for

follow-up

Conclusions The preliminary results from the Canadian CVS trial suggest that the potential difference in loss rates between CVS and amniocentesis is not more than 2-4% for women 35 years of age and over with a viable fetus at the time of procedure. Thus, among every 200 such women having CVS, there are likely to be at most 5 additional fetal losses (spontaneous and induced) than there would be among 200 women having amniocentesis. The difference is greater when all women allocated to a procedure are considered, but still not significant. A tendency to later losses in the CVS group is reflected in a higher crude perinatal mortality rate and the greatest imbalance between the two groups is the stillbirth rate after 28 weeks, for which at present there is no obvious explanation. Since the proportion ofpreterm births and small-for-dates infants was similar in both groups, as were the mean birthweights, the high late fetal loss rate after CVS may reflect chance alone. However, we are continuing to gather information on any additional late losses to see if this trend persists. Maternal morbidity was similar in both groups, and there were no maternal deaths. About 10% of women undergoing CVS required an amniocentesis to complete prenatal diagnosis. While the cytogenetic findings raised the question of false positive and false negative results, these were clarified with further investigation. In this trial we could not always decide on eligibility before randomisation so the women entering the trial and those remaining eligible after randomisation may reflect different selection processes. Nevertheless, the total loss rates are not significantly different between groups no matter whether data on all randomised or on eligible women only are analysed and, consequently, this result may reassure women on the safety of first trimester CVS. We thank the women who, generously and altruistically, took part in this study; the Medical Research Council of Canada for financial support; and our International Advisory Committee for encouragement.

Correspondence to: Prof J. L. Hamerton, Department of Human Genetics, University of Manitoba, 250 Old Basic Science Building, 770 Bannatyre Avenue, Winmpeg, Manitoba, Canada R3E OW3.

REFERENCES

N, Mohr J. Genetic diagnosis in the embryo by means of biopsy from extra-embryonic membrane. Bull Europ Soc Hum Genet 1968; 2: 23-29. 2. Jackson L CVS Latest News 1988 (August), no 26. 3. Hook EB, Cross PK, Jackson L, Pergament E, Brambati B. Maternal age specific rates of 47, +21 and other cytogenetic abnormalities diagnosed in the first trimester of pregnancy in chorionic villus biopsy specimens: comparison with rates expected from observations at amniocentesis. Am J Hum Genet 1988; 42: 797-807. 4 Vekemans M, Perry TB, Lippman A, Hamilton EF, Fournier PJR. CVS followed by amniocentesis: how often and why? Am J Human Genet 1986; 39: A266 (abstr). 5. Worton AG, Stem RA Canadian collaborative study of mosaicism in amniotic fluid cell cultures Prenat Diagn 1984; 4: 131-44 6. Kalousek DK, Dill FJ. Chromosomal mosaicism confined to the placenta in human conceptions. Science 1983, 221: 665-67 7 Watanabe M, Ito T, Yamamoto M, Watanabe G. Origin of mitotic cells of the chorionic villi in directed chromosome analysis. Hum Genet 1978; 44: 191-93. 8. Markert CL, Petters RM. Manufactured hexaparental mice show that adults are derived from three embryonic cells Science 1978; 202: 56-58 9. Crane JP, Cheung SW An embryogenic model to explain cytogenetic inconsistencies observed in chorionic villus versus fetal tissue Prenat Diagn 1988; 8: 119-29. 10. Benn PA, Hsu LYF. Maternal cell contamination of amniotic fluid cell cultures. Results of a US nationwide survey. Am J Med Genet 1983; 15: 297-305. 11. Coleman DV. Culturing trophoblast samples In: Fraccaro M, Simoni G, Brambati B, eds. First trimester fetal diagnosis. New York Springer-Verlag, 1985: 147-56. 12. Blakemore KJ, Samuleson J, Breg WR, Mahoney MJ. Maternal metaphases on direct 1. Hahneman

chromosome preparation of first trimester decidua Hum Genet 1985; 69: 380. MJJ, Perry TB. Cytogenetic analysis of chorionic villi. a technical assessment Hum Genet 1986; 72: 307-10

13 Vekemans