Significant increase in trisomy 21 in Berlin nine months after the Chernobyl reactor accident: temporal correlation or causal relation?

Trisomy 21 in West Berlin, 1980-9

Up till late 1989 the “island” of West Berlin provided a favourable setting for the epidemiological study of trisomy 21. Cytogenetic services were provided solely by the university institute of human genetics and one associated laboratory, so that all prenatal and postnatal diagnoses should have been recorded there.

This analysis is restricted to cases with standard trisomy 21 diagnosed from 1980 to 1989. All probands with translocations and mosaicism were excluded, because these conditions may have a different aetiology. Misclassification is highly unlikely in prenatally detected cases, and ascertainment therefore depended only on the numbers of women having prenatal diagnosis. Ascertainment of liveborn cases with trisomy 21, however, depends on awareness by paediatricians who examine newborn infants. In such cases the median age at tentative clinical diagnosis was 4 days, 94% of affected infants being diagnosed cytogenetically within 14 days. Cytogenetic tests were provided free during the study period, which might have aided ascertainment.

Analysis and results

Data are reported from 1980 because from that date the criteria for ascertainment remained unchanged. Moreover, during the 10 years the average age of all pregnant women remained fairly constant (around 27.4 years; table I). The same was true for the percentage of childbearing mothers aged 35 or more. This was invariably above 10%, with an average of 11.2% over the 10 years (table I).

Of the 190 073 live births registered during the study period, 226 (0.1%) were cases of trisomy 21. A further 101 cases were diagnosed prenatally (table I). There was an increasing number of prenatal diagnoses during the 10 years, from 785 in 1980 to almost 2000 in 1989. Cases diagnosed prenatally were weighted by 0.7, because only this proportion is expected to survive until birth.5 Thus an estimated total of 297 cases of trisomy 21 were diagnosed prenatally and postnatally among the live births. This high proportion (0.16% or one in 640 live births) supports the assumption of almost complete ascertainment. As a further check on ascertainment we compared the maternal age specific risk figures for trisomy 21 in Berlin with those of a Swedish study6 in which ascertainment was very high in a comparative analysis of 19 populations.7 From that study the expected number of trisomy 21 cases in our population was 290. The estimated total exceeded this and thus underlines the completeness of ascertainment.

The highest number of trisomy cases occurred in 1987 (table I). This was due almost exclusively to a peak of 12 cases in January (10 among newborn infants, two among fetuses diagnosed by chorionic villus sampling). The monthly prevalence of cases for the whole observation period did not show a consistent tendency to increase or decrease over time (figure).

Time series analysis of monthly prevalence of all prenatally and postnatally diagnosed cases with trisomy 21 in West Berling from January 1980 to December 1989 (solid line). Autoregressive moving average model that fitted data reasonably well is superimposed (broken line)

• Download figure
• Open in new tab
• Download powerpoint

Spectral analysis, which estimates the amount of variance in a series accounted for by cycles of different periods, excluded any kind of gross periodic changes. Several simple autoregressive moving average models were applied to the time series.8 These help to detect random fluctuations that exceed usual levels and determine the point at which variation is too great to attribute to chance. One model that fitted the data reasonably well (and which did not have a seasonal component)1, 2 is shown in the figure. The autocorrelations and partial autocorrelations of the residuals were randomly distributed and the box-Ljung statistics was not significant in any lag. The model also gives the upper confidence interval (99%) for the original series. Plainly the prevalence in January 1987 greatly exceeded that value. The same was found for all other autoregressive moving average models calculated. All statistical calculations were carried out with SPSS/PC+ V.3.0.1.

Trisomy 21 is among those classes of genetic defects that are easily identified and show almost no seasonal variation. Thus, in addition, we tested the goodness of fit for the monthly frequency of trisomy 21 with the expected Poisson distribution. As a Poisson distribution is defined only for integers, we did not apply the correction factor (0.7) to prenatal cases but counted every case diagnosed prenatally and postnatally as an observed case. This gave a mean of 2.7 cases per month over the 10 years. With this mean taken as (lamba) the observed cases of trisomy 21 per month fitted a Poisson distribution (P=0.99; Kolmogorov-Smivrnov test). We did not take into account the variation in total monthly births, because the number of trisomy 21 cases per month did not show any correlation with these figures in our dataset. The upper 99% confidence interval for the Poisson distribution estimated according to to Clopper-Pearson9 was 10.5 cases. The number of observed cases in January 1987 exceeded this, so making it unlikely that the temporal increase in trisomy 21 occurred by chance.10, 11

Possible causative factors Maternal age

Maternal age is the most important risk factor for trisomy 21. Of the 12 mothers in the January 1987 cluster of cases, six were aged 35 or over (table II). This agrees with the age distribution of all 327 mothers of trisomy 21 cases in the series, 148 (45%) of whom were in this age group. The mean (median) age of the 12 mothers (33.7 (34.5) years) differed only slightly from that of all mothers of cases (33.0 (34.2) years). To assess this difference 100 random samples of 12 women were taken from the whole cohort with the procedure SAMPLE of SPSS. The mean (median) age in 33% (42%) of the sample was higher than that of mothers in the January 1987 cluster, confirming that the age distribution was not in an extreme range.

Radiation exposure

Nine of the 12 parents in the January 1987 cluster were interviewed by a physician about their obstetric, medical, and radiation histories in the year before conception (table II). Two couples refused interviews and one had emigrated. No serious illness or drug misuse was reported. The mother in case 3 had had a routine chest x ray examination in May 1985, and the mothers in cases 10 and 12 had had radiography of pelvis and the teeth, respectively, in March 1986.

The only common exogenous factor was exposure to radioactive clouds from the nuclear explosion at Chernobyl which had passed over Berlin nine months previously. Only one couple was not in Berlin at the time. None of the interviewees expressed exaggerated anxiety over the exposure to radiation. Two couples, however, had begun excluding meat and milk products from their diet.

The first cloud reached Berlin on 29 April 1986 at 6 pm, leading to a large increase in airborne radioactivity, mainly due to nuclides with short half lives such as iodine-131 (eight days) and tellurium-132 and iodine-132 (78 hours). The second cloud passed over Berlin on 4 May. During the whole period the weather was sunny and dry. A heavy shower during the night of 7-8 May removed practically all artificial radionuclides from the atmosphere and subsequently any inhalation of these was negligible. Irradiation from the ground was increased by a maximum of only 60% above normal.12

The main exposure to the Berlin population was from 29 April to 8 May and was due primarily to inhalation of ¹³¹I and ¹³²I,RF 13* which accumulate preferentially in the thyroid. No representative direct measurements of the thyroid dose have been performed in Berlin. However, for this period the time integrated airborne radioactivity of particulate bound iodine and caesium in Berlin was estimated as 100-200 and 50-100 Bq/h/m³, respectively.13 This calculation excluded contributions from gaseous elemental and organically bound radioiodine, so that the total activity was likely to have been three times higher.13 With respect to the overall dose received during the period a severalfold increase above normal seems to be realistic.

Based on the date of the first day of the last menstrual period, cases 4-8 were conceived during the time of highest exposure (table II). Given the variability in the length of the preovulatory cycle and the uncertainties in determining the first day of last menstruation, cases 2,3, and 9-11 may also have been conceived during this period or shortly thereafter. The parents in case 9, however, were not in Berlin at the time (table II).

In seven of these 10 families the origin of the extra chromosome was studied by examining chromosomal heteromorphisms. In six cases failure had occurred during oogenesis and thus at around the time of conception. In case 11 paternal nondisjunction was observed, which must have occurred about two months before conception14 and consequently before the Chernobyl reactor accident. Thus in eight cases (2-8 and 10) a temporal correlation between meiotic failure and the increase in radioactivity in Berlin cannot be excluded. The question is whether this also points to a causal relation.

Iodine uptake

Indirect evidence suggests that the uptake of radioactive iodine after the Chernobyl accident may have been somewhat higher in West Germany than in Hungary and most other European countries. Iodine prophylaxis is not mandatory in West Germany, and the prevalence of iodine deficiency is among the highest in Europe.24, 25 For example, median urinary concentrations of iodine in full term newborn infants were 883 and 867 nmol/l in Helsinki and Stockholm but only 221 nmol/l in Berlin.26

We measured thyroid stimulating hormone concentrations in newborn infants as an index of maternal iodine uptake during pregnancy. In two cases with trisomy 21 a transient increase was recorded (table II). Values were, however, in the normal range for newborn infants with Down's syndrome.

As the inhalation of radioactive iodine occurred at around the time of conception a direct effect of low dose irradiation on the gonads must be taken into account. This does not, however, rule out completely an indirect mechanism acting via altered thyroid function. There are several, if controversial reports that mothers of children with Down's syndrome have a higher prevalence of thyroid antibodies and thyroid disease.27, 28 Either case would imply that low dose irradiation at the time of oogonial meiotic divisions could induce nondisjunction.

So far as we know there are only two epidemiological studies in which parents were exposed to a high dose of irradiation at the time of conception. These concerned the inhabitants of Kerala, India, and Yangjiang County, China, who received high background radiation from monazite soil containing thorium. The dosage was perhaps comparable to that in Berlin during the critical two weeks. Though some ascertainment bias could not be excluded,4 both studies found a significant increase in trisomy 21.29, 30 In conjunction with our findings, this implies that chromosomal segregation during human oogenesis can easily be disturbed.

Direct evidence that maternal meiosis may be disturbed by exogenous factors at the time of conception remains a matter of controversy, as does the postulated effect of preconceptional x irradiation or viral infections.3, 31 Recently a convincing correlation between the accidental ingestion of metriphonate (employed against fish parasites) at around the time of conception and trisomy 21 has been reported.32 Indirect evidence, however, is provided by the occasionally observed seasonality in the frequency of aneuploidies, which has been explained by seasonal variation in endocrine factors34, 35 or reproductive activity,35 and especially by the high spontaneous rate, possibly affecting 20-30% of all zygotes.36 This incidence is more than an order of magnitude higher than in mice, which is why with respect to the induction of nondisjunction (but in contrast with gene mutations) extrapolation from mice to humans poses problems in principle.

Taking all the evidence together, we conclude that the significant increase in trisomy 21 in Berlin nine months after the Chernobyl reactor accident was not simply a chance event. Assuming (a) that maternal meiosis is an error prone process that is most sensitive to endogenous or exogenous factors at around the time of chromosomal segregation, and (b) that inhalation of large amounts of radionuclides with short half lives, (especially ¹³¹I) was limited to a period of less than two weeks in Berlin, we believe that a causal relation was the most likely explanation. The extent of the increase in trisomy 21 remains difficult to explain, however, despite the low amount of radioactive fall out that people in Berlin were exposed to.

Our interpretation of the findings contradicts current textbook opinion. Accepting it as valid has immediate practical consequences for genetic counselling and also considerable theoretical implications concerning the process of chromosomal segregation in humans.

This work was supported by a grant from the Umweltbundesamt. Data on maternal age distribution were kindly provided by O Burkhardt from the Statistisches Landesamt Berlin. We are grateful to Professor Dr H Lindenberger and Professor Dr F Vogel for their most valuable comments and to Dr M Digweed for help in preparing the manuscript.