Results

STUDY POPULATION

Details of the total study population are given in table 1. There are 809 state schools in our area and 527 participated. We held clinics in 360 schools while 167 had no boys suitable for testing. Of the 282 non-participating schools, 42 declined to join the survey and 240 did not reply. There was a total of 135 798 boys aged 5 to 18 in the 809 eligible schools, while in the 527 schools which participated there were 82 294 boys in the appropriate age range. The test was offered to 5451 boys of whom 3738 were tested, an uptake of 69%. Of the 1713 boys not tested, 175 refused, 37 letters were returned, 68 agreed to testing but were absent on clinic day, and 1433 did not reply to either the original or the follow up letter. We received 2968 samples from mothers of the boys tested, an uptake of 79%. A further 244 mothers agreed to give samples but failed to return the sample pack. The remaining 527 mothers either refused or were not available for testing.

Among the 809 eligible schools, 50 were special schools in which all the pupils had either moderate learning difficulties (MLD), severe learning difficulties (SLD), physical difficulties (PD), or emotional and behavioural disorders (EBD). One third of the boys tested in this survey attended special schools (table 1) and the remaining two thirds attended regular state schools. Forty nine of the special schools participated and we held clinics in 46. Of those not visited, three had no boys suitable for testing (two schools for EBD boys and one for boys with PD) and one school (MLD) did not reply. The high uptake among the special schools is presumably a reflection of a perception that there is more likely to be a medical explanation for the learning difficulties associated with pupils at the lower end of the academic scale.

PREVALENCE OF FULL AND PREMUTATIONS

In the study population

Alleles for FRAXA and FRAXE were divided into five classes: minimal, common, intermediate, premutation, and full mutation (tables 2and 3) based on size and methylation status. As can be seen from table2 we found 20 full FRAXA mutations and two premutations with repeat numbers over 60 among the tested boys, and of these 17 full mutations and one premutation were already known to us through the diagnostic service. Among the 20 full mutations, five were mosaics, and of these, two were methylation mosaics while three had both pre- and full mutation sized alleles. Among the mothers' control chromosomes we observed no full mutations and one premutation. All the full mutations were found in boys attending special schools. As can be seen from table3, we found a single full FRAXE mutation and two premutations among the tested boys and of these the full mutation was already known to us through the diagnostic service and was attending a special school. No full or premutation sized alleles were found among the mothers' control chromosomes.

When a boy, already known to us as having a FRAXA or FRAXE mutation, appeared on a school list for testing, we did not contact the boy or his family again, but included him on our list of tested subjects. Clearly this may have inflated the number of full mutations in our tested population as a proportion of these subjects might have refused to be tested had we approached them in the absence of a diagnosis. However, we have attempted to allow for this possible source of bias when calculating the population frequency of full FRAXA and FRAXE mutations.

The failure rate of tested samples from the boys was 0.16% (six boys) at FRAXA and 0.2% (seven boys) at FRAXE (tables 2 and 3). Of the six boys for whom we failed to obtain a result at FRAXA, the mothers of three were heterozygous for two common alleles and in two out of the three remaining cases the DNA failed for all tested loci. Therefore, for five boys the most likely explanation for failure was insufficient or degraded DNA. We are concerned about the one remaining case, for whom the FRAXA result is unresolved, because we obtained results for all markers except FRAXA. Therefore, he could be either a pre- or full mutation that we did not detect using PCR methods. However, we have been unable to obtain a repeat sample from this boy. For the seven boys who failed at FRAXE, we obtained heterozygous results in the normal range for five mothers and the two remaining cases failed for all markers, so the most likely explanation for failure in all seven cases is technical. The failure rate for the mothers was 1.2% at FRAXA and 1.1% at FRAXE. The higher failure rate among the mothers was probably because they took their own samples, whereas the samples from boys were obtained by an experienced genetic nurse.

In the total population

There were three groups of boys in the population who were not tested, and who were at different risks of having a full mutation: (1) eligible boys in special schools, (2) eligible boys in regular schools, and (3) boys who were not offered the test. Although no FRAXA or FRAXE subject with a full mutation is known to the diagnostic laboratory among boys not eligible for the test, there may be one or more among boys not yet recognised as having SEN or recognised but not tested. In the absence of a generally accepted method to allow for these contingencies, we used the Poisson distribution15 to estimate a 95% confidence limit for the number of cases missed in those two groups as m=−ln 0.05=3. Within the FRAXA group with 20 cases, the 95% Poisson confidence interval is 12.22-30.89. Summing the two risk groups and dividing by the school population, a conservative interval is 1/8922 to 1/4007 with midpoint 1/5530. An alternative calculation that neglects sampling error in the observed number is 23/135798=1/5094. If, in addition, the possibility of missed cases is ignored, the estimate is 1/6790. Both are well within the Poisson confidence interval. A similar calculation for FRAXE full mutations among males is 1/23 423 (confidence interval 1/44 887-1/15 486).

If we assume that premutation sized alleles have no phenotypic effect with respect to intellectual performance, the frequency in the studied population, both boys and the mothers' control chromosome, is the population prevalence. We tested a total of 6664 chromosomes for FRAXA and found three to have a premutation with between 61 and 200 repeats, a frequency of 1 in 2221. However, if under a less stringent model we regard alleles of 51 to 200 repeats as premutations, the frequency becomes 1 in 208. As the definition of a premutation is arbitrary, it is not possible to determine its population prevalence with accuracy. If we assume that premutation and intermediate alleles have a deleterious effect on intelligence (see below), the prevalence is given by the control chromosomes alone, namely 1 in 2932 for alleles over 60 and 1 in 418 for alleles greater than 50.

Similarly for FRAXE, we tested a total of 6667 chromosomes and found two to have a premutation with repeat size over 60, a frequency of 1 in 3334, and seven to have a repeat size over 40, a frequency of 1 in 952. If we assume premutation and intermediate alleles have an effect on intelligence, the prevalence for alleles over 60 repeats is extremely low while that for alleles over 40 repeats is 1 in 2936.

FREQUENCY OF INTERMEDIATE ALLELES

The distribution of allele sizes is given in tables 2 and 3. When we analysed our data after the first two years of the study we were surprised to find a significant excess of intermediate and premutation sized alleles for both FRAXA and FRAXE among the boys by comparison with the controls.10 We have now analysed the frequency of intermediate alleles for FRAXA and FRAXE in the whole population. For this analysis, all XXY boys and their mothers, as well as one XXX mother, were removed from the samples.

If we treat our study population as two separate cohorts, the first two years when we initially found the unexpected result10compared to the final three years which can be regarded as an independent population, we see from table 4 that the excess of intermediate and premutation alleles is significant in both cohorts for FRAXA, lending independent evidence to the hypothesis that intermediate and premutation alleles have an effect on intellectual ability. For FRAXE, the result is inconclusive as significance for the second sample has diminished. However, the pooled data remain significant. The excess of intermediates for FRAXA was similar in the whole study population and in the special schools subpopulation, suggesting that the effect of intermediate alleles is apparent across a wide range of learning difficulties.

STABILITY

We tested 2932 mother to boy transmissions and found only eight changes in transmission in the common and intermediate range, five at FRAXA and three at FRAXE (table 5). In every case the haplotype was known from family studies. We have also considered the transmissions from premutations. There were 16 for FRAXA (table 5), the remaining six cases being from mothers with a full mutation. At FRAXE there were two transmissions from mothers with premutations, and none from full mutations.

BROTHERS

Among the survey boys, we tested 160 pairs of brothers for whom we had maternal data, which included 15 pairs of twins. All twins were excluded except for two sets who were identical for all loci tested and had a brother; these twins versus the singleton brother were counted as one pair. This resulted in 147 sib pairs eligible for analysis. Each pair was examined in order to determine if among these brothers with SEN there was a tendency to inherit the same maternal haplotype as defined by the tested polymorphic markers detailed in the Methods section. However, no significant tendency was seen, as 80 pairs of brothers inherited the same maternal haplotype and 67 pairs inherited different maternal haplotypes for the six loci tested (χ²=1.15). These results suggest there is not a gene closely linked to the Xq27.3 region which contributes significantly to the special educational needs of the brothers in our population.

KLINEFELTER SYNDROME

During the survey we diagnosed 15 boys, including one pair of twins, who had two alleles at a number of the X loci we tested. In every case a blood sample was obtained and a 47,XXY karyotype found on standard cytogenetic investigation. Six of the boys had inherited both maternal alleles suggesting a maternal origin for both X chromosomes, while seven, including the twins, had inherited one maternal allele and at some loci an allele that was not present in the mother, so must be paternal in origin. In two patients there was no maternal sample available, so the origin of the additional X is not known. Klinefelter syndrome is present in 1 in 1000 newborn males and is associated with phenotypic abnormalities including a reduction in mean IQ.16 As a result it is found with increased frequency among males with mental impairment, although its frequency in a population of boys with SEN is not known. We detected 15 cases in the 3738 boys studied (0.4%), but our methodology is likely to detect only a proportion of 47,XXY males. It is known that approximately 50% are of maternal and 50% of paternal origin.17 We tested six highly polymorphic loci in Xq27-28 and so should have detected most of the paternal cases. We are likely to have missed the Klinefelter cases who are homozygous for the markers tested, either as a result of a maternal meiotic non-disjunctional error associated with recombination or a postzygotic mitotic error. Crawford et al,18 in a similar survey, detected seven 47,XXY males among 1979 tested (0.35%), a frequency very similar to that seen by us. In contrast to our observation of only three previously undiagnosed full FRAXA mutations and no previously undiagnosed FRAXE mutations among the 3738 boys tested, 13 of the 15 with Klinefelter syndrome were previously undiagnosed. While we are only likely to have detected a proportion of such cases, it was a somewhat unexpected bonus, as it provided an explanation for the SEN in these boys.

Discussion

Since the introduction of a molecular technique for the diagnosis of FRAXA expansions, there have been a number of studies aimed at determining the incidence of full mutations in the population. The great majority of these were based on surveys of mentally retarded males and subsequent extrapolation to give a population incidence of FRAXA full mutations. Most are based on surveys of between 200 and 2000 subjects and estimate a population frequency of approximately 1 in 4000.5 6 18 19 These include a molecular reappraisal of the cytogenetic surveys of Turner et al 4 and Webb et al,3 and in both cases the frequencies given by cytogenetic studies were reduced substantially.

Recently de Vries et al 20published a large survey in the Dutch population in which they found a prevalence of approximately 1 in 6000 full FRAXA mutations, a figure very similar to the prevalence of 1 in 5500 in our survey, while Crawford et al 18 in a somewhat smaller survey found a prevalence of 1 in 3460 for the general white male population and 1 in 4048 for the African-American male population. However, the 95% confidence intervals for these estimates include the figures estimated by de Vries et al 20 and by ourselves.

In any survey that relies on informed consent, it is always possible that because of a personality trait associated with the genotype under study, the condition occurs with a higher frequency among those who did not agree to participate than among those who did. The reverse situation could also occur. This potential source of bias is difficult to detect and thus impossible to correct. However, it seems reasonable to suppose, based on the weight of evidence currently available, that the population prevalence of FRAXA full mutations is between 1 in 4000 and 1 in 6000.

It is of interest that in our survey we found only three of the 20 FRAXA full mutations to be previously undiagnosed cases, while the remaining 17 were already known to us. This led us to conclude that in our jurisdiction, which has had excellent diagnostic services for some 30 years, the great majority of fragile X full mutations have already been diagnosed. This contrasts with the survey of de Vrieset al 20 in which 50% of full mutations were previously undiagnosed.

An estimate for FRAXE prevalence of 1 in 50 000 has been suggested.21 However, this was not based on direct testing. There are only two large population based surveys in which an estimate of the prevalence of FRAXE full mutations can be based, namely that of Crawford et al 18 and the present report, and it is clear that it is extremely rare. Crawfordet al 18 found none among 1866 subjects while we found only 1 among 3731 boys with SEN, giving an estimated population prevalence of 1 in 23 500. Numerous other small surveys of mentally impaired subjects have been reported19 22-26 and in a total of 3769 tested subjects only two had FRAXE full mutations. Because of the rarity of FRAXE full and premutations, there is no clear description of the associated phenotype. While it is generally agreed that expansions are associated with mild mental retardation in the absence of physical abnormality, the relationship between the retardation and the expansions is not at all clear.

We reported an unexpected excess of intermediate and premutation alleles among the boys in our survey for both FRAXA and FRAXE in the first two years of the study.10 We continued to show this excess for FRAXA in the last three years of the survey and over the whole study the excess is significant at the 0.001 level. In contrast, the excess shown for FRAXE in the first two years was not seen in the last three years of the survey, although the excess overall is still significant but only at the 0.03 level. As our controls were the X chromosomes of the mothers that were not transmitted to the SEN sons, one formal explanation for our observations would be the preferential transmission of the larger maternal allele to the son. We tested this and found no evidence for any preferential transmission either in the survey population or among the families in which intermediate, pre-, or full mutations were segregating.27 Thus we conclude that the excess of intermediates and premutations for FRAXA among the boys with SEN may be real and that intermediate and premutation alleles have a deleterious effect on intellectual performance or behaviour or both. There is now a well established phenotypic effect of premutation sized alleles, namely premature ovarian failure,28 so our observation of a second biological effect of large unmethylated repeats is not without precedent. Unfortunately, we are unable to obtain phenotypic information on the boys with intermediate alleles and therefore are unable at present to define their phenotype.

Since we first published our observations of an excess of intermediate and premutation sized alleles among the boys with SEN, at least four studies have addressed the same problem. In three studies which included White American, Black American, Cypriot, Canadian, and French cases,18 19 29 no increased frequency of intermediate or premutation alleles was found among retarded males by comparison with the appropriate controls. In the remaining study, carried out in Brazil on relatively small numbers of patients and controls, there was an increased frequency of intermediate and premutation alleles among the impaired males, but this did not reach statistical significance.30 In the absence of phenotypic data from our population, it is impossible to know whether the differences among studies are the result of different ascertainment criteria or attributable to some other cause. If our results on FRAXA are valid they suggest that the contribution to mental impairment of intermediate and premutation alleles is greater than that of full mutations. While the attributable penetrance of mental retardation among full mutations is 1 and that for intermediate/premutations is much less, the number of the latter is much greater.27

As we were particularly interested in the stability of transmission of the FRAXA and FRAXE repeats from mother to son, we separated PCR products on an ABI 377 sequencer so that even instabilities of one repeat size could be reliably detected. Among 3020 transmissions for FRAXA we found 21 mutations, of which only one was in the common range and four in the intermediate range. Thus, transmissions of FRAXA repeats, at least from females, seem remarkably stable, although instabilities are approximately 90 times more frequent in the intermediate than the common range. Among 3054 transmissions of FRAXE we found five mutations, three in the intermediate range and two in the premutation range, again showing how stable such transmissions are, especially in the common range.

In conclusion, it is of interest that all full mutations detected by us were in boys attending special schools, although they formed only one third of all boys tested. Thus, it is clear that in our population FRAXA and FRAXE full mutations are extremely rare among boys with SEN who attend ordinary schools. Our data suggest that population screening for FRAXA and FRAXE would not be cost effective in a population which has been in receipt of comprehensive genetic services for a considerable period. However, in a population in which a significant proportion of FRAXA mutations are undetected, it could be cost effective to screen a population of boys in special schools, selected on the basis of phenotypic features associated with the FRAXA syndrome, as this is known to make such screening considerably more effective.31