Introduction

Estimates of the incidence of chromosomal abnormalities at birth are derived largely from studies in newborns carried out in the 70s, before the widespread use of prenatal diagnosis and pregnancy intervention. In 1977, Hook and Hamerton1 reported on 56 902 newborns and subsequently Benn and Hsu2 looked at 68 159 cases collected in the same years to infer the frequencies of the major chromosomal abnormalities. These data have not been further updated and because the early studies had technical and methodological limits in detecting some structural abnormalities and mosaicisms, our knowledge of the frequencies of those abnormalities in the general population has remained scanty. Data on 52 965 amniocenteses have been used to calculate the frequencies of specific abnormalities in the second trimester of pregnancy in cases with maternal age >35 years, estimating the age-dependent risks for the major aneuploidies;3 moreover, Hook et al4 showed that maternal age-specific rates of clinically significant cytogenetic abnormalities may be used in conjunction with the data on the natural history of chromosomally abnormal fetuses to infer maternal age-specific rates of cytogenetic abnormalities in live-born infants. Therefore, from the information gathered at the second trimester of pregnancy, it is possible also to establish the overall incidence of the chromosomal abnormalities in live-born infants.

Recently, the increasing reliability of non-invasive techniques has facilitated far more first trimester screening for the major aneuploidies5 and in the European cities genetic counseling has led to more judicious use of amniocentesis for prenatal diagnosis.6 However, the selection of cases for cytogenetic analysis creates a considerable ascertainment bias for estimates of the frequency of many chromosomal abnormalities in newborns in the general population.

To circumvent the ascertainment bias, we have focused on data from 1995–1996. In that era the impact of the first trimester screening strategies was still negligible, and we assess here the incidence of non-age-dependent chromosomal abnormalities in the second trimester of pregnancy. In Italy national genetic data, including the number of cytogenetic analyses carried out in each laboratory, have been collected since 1987.7 For 1995–1996, data also included the results of cytogenetic analyses, grouping diagnoses into three categories: women ≥35 years, women <35 years and women who underwent amniocentesis based on familial, ultrasonographic or biochemical indications.8

We have compared the frequencies of the chromosomal abnormalities in 51 758 women ≥35 years with those in 37 207 women <35 years, and have then evaluated the age-dependency of single chromosomal abnormalities. For the non-age-dependent abnormalities, frequencies in the second trimester of pregnancy were established across the entire study cohort.

Materials and methods

Data accrual

Supported by the Italian National Institute of Health (Istituto Superiore di Sanità), a retrospective study was carried out by accessing the results of cytogenetic analyses of amniotic fluid performed in Italy during the period 1 January 1995–31 December 1996. Data were collected from the records of the public and private laboratories performing prenatal cytogenetic diagnosis in Italy. The laboratories had been earlier identified by the Italian Association of Medical Cytogenetics (AICM) and adhered to the AICM guidelines,9 requiring banded chromosomes (G or Q banding) with a resolution above the 320-band stage and a minimum cell count, according to Hsu et al.10 A form was distributed to each laboratory to record the following data: the initials of the proband, age of the mother (< or >35 years, calculated at the time of amniocentesis), the result of the karyotype analysis (described according to the International System for Human Cytogenetics Nomenclature – ISCN-198511) and the indication for undergoing the examination.

Of the 106 laboratories operating in 1995 in Italy, according to the AICM guidelines, 98 (92%) participated in the study. Two laboratories did not specify items in the standard form and were thus excluded from the analysis; six laboratories did not respond.

A total of 98 118 amniocenteses were accessed: among them, 2025 cases were excluded because of poor/incomplete reports or because amniotic fluid analysis was carried out after the finding of an abnormality in chorionic villi analysis and 7128 cases were excluded from the analysis because there was a biochemical, ultrasonographic or familial indication for the analysis. In the latter exclusion group, those women undergoing amniocentesis with maternal age ≥35 years and a clinical, biochemical or familial indication were also included.

All the chromosomal abnormalities detected by karyotype analysis, including the mosaics, were classified into the following categories: (a) autosomal aneuploidies, (b) sex chromosomal abnormalities in males, (c) sex chromosomal abnormalities in females, (d) polyploidies (e) balanced autosomal rearrangements and (f) unbalanced autosomal rearrangements.

The cytogenetic nomenclature adopted for the definition of abnormal karyotype was classified according to the criteria reported in ISCN 1995.12 All chromosomal variants, defined following the ISCN 1995, were excluded.

Data analysis

Rates of sex chromosomal abnormalities were calculated for the affected sex, assuming equal numbers of males and females in the prenatal series. Statistical analyses were carried out by comparing the frequencies of chromosomal abnormalities at amniocentesis, using χ2 analysis. Owing to the large number of comparisons carried out, to reduce the risk of a Type 1 statistical error, significance was set at P<0.01.

Results

This study thus reports the results of 88 965 amniocenteses with a total of 1607 chromosomal abnormalities, including 224 mosaic. The frequencies of chromosomal abnormalities observed in second trimester amniocenteses for women ≥35 years were compared with the frequencies in women <35 years of age (Table 1). The trisomies 21 and 18 and the 47,XXX and 47,XXY karyotypes were largely age-related (P<0.001); trisomy 13 showed a significant age-dependency (P<0.01), whereas the 47,XYY, 45,X and triploid karyotypes showed a borderline significance (P=0.01). On the other hand, among the autosomal mosaic aneuploidies, trisomy 21 showed a significant age-related increase (P<0.01), suggesting that the incidence of the mosaicisms, with the exception of trisomy 21, is not significantly influenced by the age of the mother.

Table 1 Comparison between the frequency of chromosomal abnormalities at the second trimester amniocentesis in women aged ≥35 and <35 years

In the category of mos 45,X, all the sex chromosome complements in male fetuses were evenly distributed in women aged above and <35 years, including the 45,X/46,XY and idic(Y) karyotypes. On the contrary, the 45,X/46,XX karyotype was significantly more frequent in the group of women >35 years of age, whereas the 45,X/47,XXX karyotype was not influenced by the age of the mother. Also chi 46XX/46XY did not show any age-dependency.

The autosomal trisomies, involving chromosomes 2, 8, 12, 22 and mosaic tetraploidy, detected in women with advanced matemal age (equal or older than 35 years) could not be compared, not being present in the group of women <35 years of age.

Both balanced and unbalanced structural rearrangements showed no significant differences between the two age groups, suggesting that the incidence of these anomalies is not significantly influenced by the age of the mother.

For specific chromosomal abnormalities that showed no significant difference with age, frequencies were calculated for the entire cohort (Table 2). Non-age-dependent mosaic autosomal and sex chromosome aneuploidies occurred in 1 out of 2600 and 1 out of 1000 amniocenteses, respectively. Trisomy 20 mosaicism and 45X/46XY karyotypes were particularly frequent, being found in 1 out of 5200 and in 1 out of 1900 amniocenteses. The frequency of chimeras is around 1 out of 18 000 amniocenteses. The total incidence of balanced autosomal rearrangements ranged from 1 out of 560 for reciprocal translocations to 1 out of 1100–1200 for inversions and robertsonian translocations, respectively. The incidence of unbalanced rearrangements was considerably lower, 1 out of 5900 for robertsonian and 1 out of 6300–6800 for derivatives and deletions (the most frequent of the latter being the terminal deletion of the short arm of chromosome 4, more frequent than 5p deletion). Ring chromosomes showed an incidence of about 1 out of 22 000, whereas supernumerary markers were scored in 1 out of 1200 cases.

Table 2 Non-age-dependent chromosomal abnormalities in the overall study population

Discussion

Major efforts are now targeted to reduce the incidence of the major aneuploidies through prenatal diagnosis, but the incidence and the impact of other abnormalities are generally more obscure. Important data have been collected on the risk for structural rearrangements, marker chromosomes, mosaicisms, etc; however, those data customarily come from selected series (ie, cases at risk for a specific abnormality)13 and can not be used to calculate the true incidence of the single rare abnormality.

On the other hand, data on incidence are vital not only for genetic counseling, but also for informed discussions of the possible desirability of focusing only on the major aneuploidies by adopting faster and cheaper new techniques of prenatal diagnosis, like QF-PCR.14 They are also important to establish a standard reference incidence to compare increased risks in procedures like in vitro fertilization.15

Our study can contribute to valid estimates of the frequency of the chromosomal abnormalities because it is based on 88 965 amniocenteses, with high compliance (92%) in Italian laboratories and uniform coverage of the population over a two-year period (1995–1996). To our knowledge, similar data are not available for other countries.

The comparison of our data with the series published by Ferguson-Smith et al3 confirms the earlier estimates for the main categories of abnormalities with surprising accuracy, reinforcing the validity of the inferences (Table 3). As shown there, apart from the overall rate of mosaic abnormalities (P<0.001), trisomy 21 was the only category that showed some indication of a difference (borderline P-value of 0.013). In addition to the data about the women >35 years of age, our study also reports the results for younger women undergoing amniocentesis. By comparing the frequencies in the two groups (< and >35 years) we could detect some age-dependency of the abnormalities. Our results confirm the age-dependency of the common autosomal and sex chromosomal aneuplodies, show a borderline significance of the age-dependency of 47,XYY, 45,X and triploidy and, surprisingly, show a significant age-dependency of the mosaicisms of trisomy 21, substantiating the hypothesis that a meiotic non-disjunction mechanism followed by rescue of trisomy could account for some of the mosaic cases.16

Table 3 Comparison between our results and those reported by Ferguson-Smith and Yates, 1984

Our study shows that the incidence of the 45,X/46,XX karyotype is significantly age-related, whereas the 45,X/47,XXX is not. This new finding can be explained with the evidence that the 45,X/47,XXX cases are derived from a mitotic non-disjunction in a normal disomic cell line, whereas the majority of the 45,X/46,XX cases almost always result from the loss of a chromosome from a normal disomic fertilization.16 Therefore, it can be hypothesized that the 45,X/46,XX karyotype originates from an anaphase lag of the X chromosome in XX zygotes because of the aging of the oocyte.

On the other hand, our results show the absence of an effect of the maternal age on the frequency of the 45,X/46,XY karyotype. Molecular studies of 45,X karyotypes show that the single X chromosome is maternally derived in about 80% of the cases, probably due to a higher risk of mispairing between the X and Y chromosomes during male meiosis, as well as a greater tendency to mitotic loss of a Y chromosome compared with an X chromosome.17 Therefore, the postzygotic loss of the Y chromosome as a mechanism of the 45,X/46,XY may well explain why this karyotype is not influenced by the age of the mother.

The level of reciprocal balanced translocations, about 1 out of 560, is consistent with other estimates from prenatal diagnosis.18 The frequency we find is about double of that observed in studies of live-born children in the 1970's,2 but is in agreement with the frequency found by Jacobs et al19 in an unselected series of newborns using moderate levels of banding.

After excluding variant forms, inversions (not scored for peri- vs para-centric) were found in about 1 out of 1100 cases, a level similar to the upper limit reported by Van Dyke et al,18 but significantly higher than the 1 out of 5200 reported by Benn and Hsu2 from historical data. By contrast, we observe a frequency of robertsonian translocations very comparable to the earlier estimate (1 out of 1186 vs 1 out of 1099 in the newborns studied by Benn and Hsu).

A mosaic abnormality was found in, approximately, 1 out of 400 amniocenteses. In particular, there were 34 non-age-dependent mosaic aneuploidies (approximately, 1 out of 2600 amniocenteses), with the most frequent being trisomy 20. The overall rate of mosaic abnormalities was significantly higher than the rate earlier reported by Ferguson-Smith et al,3 consistent with the improvement in current diagnostic guidelines for mosaicism in prenatal diagnosis.10 The finding of a high frequency of mosaic trisomies may be taken into account when considering the causes of unexplained phenotypes, as the chromosomal mosaicism could be involved in the pathogenesis of several disorders.20

The overall frequency of supernumerary markers, about 1 out of 1200, is close to the upper limit of the reported range in earlier studies.4 No information was available to permit us to distinguish between de novo and familial markers, and this probably explains the absence of an effect of maternal age that has been reported in de novo cases.13 A recent metanalysis of the frequency of small supernumerary marker chromosomes in prenatal and postnatal cases reported respective incidences of 0.075% (1 out of 1333, very close to ours) and 0.044% (1 out of 2270).21

The overall frequency of deletions that we observed (1 out of 6800) is similar to the earlier estimate by Benn and Hsu in a metanalysis of studies of newborns.2 In both, the most frequent deletion is of terminal 4p, with a frequency of 1 out of 21 000 and only 4q34-qter, 5p and 18q11 deletions showed frequencies higher than 1 out of 50 000 – each reported in two cases. All other deletions were found only once in the entire cohort. The numbers are small, but it is of interest that the observed frequency of 4p abnormality is double of that reported in another study in which the cases were clinically ascertained,22 whereas the 5p deletion (1 out of 44 000) falls into the reported frequency of 1 out of 20 000 to 1 out of 50 000.23

In agreement with an earlier study, duplications were detected less frequently than deletions, in about 1 out of 10 000 cases, even though both are suggested to be caused by the same mechanism of misalignment and unequal crossing over mediated by the presence of low-copy repeats.24 However, because duplications are more difficult to see with conventional cytogenetic techniques, an ascertainment bias may artificially lower the apparent ratio of duplications to deletions.

In conclusion, by analyzing an unselected series from a country-wide population catchment, we have inferred the incidence of a range of chromosomal abnormalities at the second trimester of pregnancy. By having a large number of cases in both groups of pregnant women, > and <35 years of age, we could determine the age-dependency of the chromosomal abnormalities, confirm most of the aneuploidies and identify some age-dependent categories like mosaic trisomy 21 and 45,X/46,XX that until now had escaped the identification. Moreover, on a series of 88 965 cases, we established the frequency of the non-age-dependent abnormalities. The data provide a reference standard for prenatal counseling and for discussions of choices for health care policy.