/1 LI 1 e - of Medical Genetics, 1977, 14, 1-9 Colchester revisiteA.l A genetic study of mental defect* NEWTON E. MORTON, D. C. RAO, HELEN LANG-BROWN, CHARLES J. MACLEAN, ROBERT D. BART, and RUTH LEW From Population Genetics Laboratory, University of Hawaii, Honolulu, Hawaii; The Kennedy-Galton Centre for Clinical Genetics, Harperbury Hospital, U.K.; and Pacific Institute of Rehabilitation Medicine, Honolulu, Hawaii summmRy This reanalysis of a classic survey leads to inferences about design of genetic studies, resolu- tion of heterogeneity, and the role of autosomal and sex-linked genes in mental retardation, which is no longer refractory to segregation analysis. By discriminating between sociofamilial and biological types we estimate that at least 351 autosomal loci can produce mental retardation, with an inbred load of 0.83 detrimental equivalents and a mutation rate of 0.008 per gamete, or less than 2.4 x 10-5 per locus. The distribution of probands was estimated as: 7 per cent medical, 60 per cent sociofamilial, and 33 per cent biological. Simple genetic mechanisms account for virtually all the biological category. Within the sociofamilial group cultural inheritance and polygenes could not be resolved. In 1938 L. S. Penrose published 'A clinical and genetic study of 1280 cases of mental defect' from the Royal Eastern Counties Institution, Colchester, England. In the preface, the Medical Research Council predicted 'There can be no doubt that Dr Penrose's methods, data, and results will to a large extent determine the general course of research in mental deficiency in this country for some years to come.' Although this anticipation proved conservative, genetic inferences were limited not only by the con- temporary state of biological, clinical, and cyto- genetic approaches, but also by restriction of genetic analysis to simple Mendelian entities. It was not until 1965 that Dewey et al were able to apply ad- vances in segregation and consanguinity analysis to part of the Colchester study, 'yielding information about the number of loci and gene frequencies of recessive factors for severe mental defect, their penetrance and mutation rates, and the mechanism which maintains these genes in the population despite the greatly reduced fertility of affected persons'. Their analysis depended on segregation of family members into severely affected and nor- mal, in frequencies that indicated a mixture of sporadic cases and recessive genes of high pene- trance. They considered that 'overlap with normal makes the group of mild defectives unsuitable for segregation analysis'. The last decade has seen extension of methods of family analysis under a model which includes in- complete ascertainment, a major locus, polygenes, and environment common to sibs, so that it is un- necessary to assume discrete segregation of normal and affected sibs (Morton, 1974). Mild mental defect is no longer refractory to genetic analysis, which we shall here apply to the Colchester exem- plar. Prevalence and incidence Penrose recognized 6 intellectual grades for patients on the basis of Stanford-Binet and Porteus tests, while their relatives were rated by interview. 'The scale of assessment was intended to corres- pond with the grades into which the patients were divided' (Penrose, 1938, p. 18). For our analysis unclassified parents were taken as of unknown score, drawn randomly from the population, while unclassified sibs were omitted (Table I). Because grades are discrete we shall not regard them as a quantitative trait. 1 * This work was supported by Grants GM 17173 and HL 16774 from the U.S. National Institutes of Health. Received for publication 12 January 1976 on October 21, 2021 by guest. Protected by copyright. http://jmg.bmj.com/ J Med Genet: first published as 10.1136/jmg.14.1.1 on 1 February 1977. Downloaded from
9
Embed
Colchester revisiteA.l genetic study ofmental defect*
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
/1 LI1 e - of Medical Genetics, 1977, 14, 1-9
Colchester revisiteA.lA genetic study of mental defect*NEWTON E. MORTON, D. C. RAO, HELEN LANG-BROWN, CHARLES J. MACLEAN,ROBERT D. BART, and RUTH LEWFrom Population Genetics Laboratory, University of Hawaii, Honolulu, Hawaii; The Kennedy-Galton Centre forClinical Genetics, Harperbury Hospital, U.K.; and Pacific Institute of Rehabilitation Medicine, Honolulu, Hawaii
summmRy This reanalysis of a classic survey leads to inferences about design of genetic studies, resolu-tion of heterogeneity, and the role of autosomal and sex-linked genes in mental retardation, which isno longer refractory to segregation analysis. By discriminating between sociofamilial and biologicaltypes we estimate that at least 351 autosomal loci can produce mental retardation, with an inbred loadof0.83 detrimental equivalents and a mutation rate of0.008 per gamete, or less than 2.4 x 10-5 per locus.The distribution of probands was estimated as: 7 per cent medical, 60 per cent sociofamilial, and 33per cent biological. Simple genetic mechanisms account for virtually all the biological category.Within the sociofamilial group cultural inheritance and polygenes could not be resolved.
In 1938 L. S. Penrose published 'A clinical andgenetic study of 1280 cases of mental defect' fromthe Royal Eastern Counties Institution, Colchester,England. In the preface, the Medical ResearchCouncil predicted 'There can be no doubt that DrPenrose's methods, data, and results will to a largeextent determine the general course of research inmental deficiency in this country for some years tocome.'Although this anticipation proved conservative,
genetic inferences were limited not only by the con-temporary state of biological, clinical, and cyto-genetic approaches, but also by restriction of geneticanalysis to simple Mendelian entities. It was notuntil 1965 that Dewey et al were able to apply ad-vances in segregation and consanguinity analysis topart of the Colchester study, 'yielding informationabout the number of loci and gene frequencies ofrecessive factors for severe mental defect, theirpenetrance and mutation rates, and the mechanismwhich maintains these genes in the populationdespite the greatly reduced fertility of affectedpersons'. Their analysis depended on segregationof family members into severely affected and nor-
mal, in frequencies that indicated a mixture ofsporadic cases and recessive genes of high pene-trance. They considered that 'overlap with normalmakes the group of mild defectives unsuitable forsegregation analysis'.The last decade has seen extension of methods of
family analysis under a model which includes in-complete ascertainment, a major locus, polygenes,and environment common to sibs, so that it is un-necessary to assume discrete segregation of normaland affected sibs (Morton, 1974). Mild mentaldefect is no longer refractory to genetic analysis,which we shall here apply to the Colchester exem-plar.
Prevalence and incidencePenrose recognized 6 intellectual grades forpatients on the basis of Stanford-Binet and Porteustests, while their relatives were rated by interview.'The scale of assessment was intended to corres-pond with the grades into which the patients weredivided' (Penrose, 1938, p. 18). For our analysisunclassified parents were taken as of unknownscore, drawn randomly from the population, whileunclassified sibs were omitted (Table I). Becausegrades are discrete we shall not regard them as aquantitative trait.
1
* This work was supported by Grants GM 17173 and HL 16774from the U.S. National Institutes of Health.Received for publication 12 January 1976
Grade Code I Q No of Parents of Sibs of More Remote Hospitalization Prevalencerange Patients Patients Patients Relatives Rate
Superior S 115 + - 11 68 82 _Normal N 85-114 - 1908 3645 10138 - -Dull Mo 70-84 179 351 493 423 } 0.03 24 492Simpleton M, 50-69 448 191 272 181 JImbecile M2 20-49 433 4 117 50 0.536Idiot Ma <20 220 - 50 14 0.25 3061
Total 1280 2465 4645 10 888 - 27 553+ 95 unknownl
Prevalence is defined as the frequency of casesliving and certifiable in the general population.There were two million persons resident in theEastern counties (Essex, Suffolk, and Cambridge-shire). Penrose (1963, p. 64) considered that only25% of severe defectives and 3% of mild defectiveswere in hospital. Since the 1280 Colchesterpatients were a representative sample of 1500patients (Penrose, 1938, p. 7), the estimated pre-valence of mild defect is (1500) (627)/(1280) (0.03) =24 492 cases in a population of two million. Forsevere defect the prevalence is (1500) (653)/(1280)(0.25) = 3061 cases. Therefore in a population of100 000 we expect 153 cases of severe mental defectand 1225 cases of mild defect.Whereas prevalence considers only living certi-
fiable cases, incidence is defined as the frequencyamong survivors to age 1 of persons who, at anytime in their life, are certifiably retarded. Diag-nosis of severe defect is virtually complete, butpremature mortality reduces prevalence. Survivalof mild defect is hardly impaired but diagnosis isincomplete, especially before entering and afterleaving school. In both mild and severe defect theratio of incidence to prevalence is roughly 2, or anincidence of 306 severe defectives and 2450 potentialmild defectives per 100 000 survivors of infancy.Tarjan et al (1973) were led by somewhat differentconsiderations to similar estimates, which, however,reflect the current tendency to diagnose severemental defect in moribund young children whopreviously were not committed to institutions forthe retarded. In Colchester, for example, no patient
TABLE IIPREVALENCE AND INCIDENCE IN A POPULATION OF
100 000
Prevalence Incidence In HospitalSource
Severe Mild Severe Mild Severe Mild
Colchester 153 1225 306 2450 38 37Tarjan et al (1973) 250 750 500 2500 - -
was below age 4 (Penrose, 1938, p. 15). At theopposite extreme of mild defect there is growingpressure for community placement or discharge andfor consideration of dull individuals, regardless oftheir adaptive behaviour, as not retarded. Thesedevelopments must affect prevalence, incidence, andhospitalization rates (Table II).
Resolution of heterogeneityWhile mental retardation is highly heterogeneous,practical considerations force us to limit geneticanalysis to a small number of categories. We de-cided to aim for three aetiological groups, based on asuggestion by Tarjan (1970):
I: BiologicalSignificant physical or laboratory pathology;
cause of retardation apparently intrinsic; diagnosisusually made before school entrance; often abnor-mal in physical appearance; mortality greater thanthe general population; retardation moderate orsevere; parental IQ and socioeconomic status repre-sentative of the general population; low birthweight,postpartum distress, parental consanguinity, orraised parental age in a proportion of cases; parentsrarely, sibs sometimes retarded.
II: MedicalSignificant physical pathology; a specific, single,
extrinsic aetiological agent probably accounts for thecondition; diagnosis made not long after environ-mental insult; abnormal in physical appearance;mortality considerably greater than the generalpopulation; retardation severe, either stationary orprogressive; parental IQ and socioeconomic statusrepresentative ofthe general population; parents andsibs rarely retarded.
III: SociofamilialCurrently available biomedical technology cannot
show significant physical or laboratory pathology;no single, specific aetiological agent can be made
Colchester revisited: A genetic study of mental defect
convincingly accountable for the condition; diag-nosis usually made after school entrance, and oftenretracted at young adult age; normal in physicalappearance; mortality not much greater than thegeneral population; retardation mild, with the dif-ference between mean parental IQ and that of thepatient frequently small; often from the socially,economically, and educationally underprivilegedstrata of our society.
Biological and medical categories are consideredto depend on megaphenic effects (i.e. large relative tothe standard deviation of IQ so that each case isprimarily due to a single, specific etiological agent),while the sociofamilial category is microphenic (dueto cumulative effects of small unidentifiable causes).Such Platonic entities are indispensable for epi-demiology, but they do not allow for dual causationby megaphenic and microphenic effects (for ex-
ample, greater risk of cerebral infection or birthinjury in the lowest social class, or predisposition ofa particular genotype to meningitis), and they do notprovide an unambiguous classification ofindividuals.Rather than a subjective classification, it seemsbetter to use a discriminant from less to more mega-phenic.A continuous discriminant can be developed from
the following considerations. Penrose (1963) esti-mated that 99% of mild defect but only 13% ofsevere defect is part of the Gaussian curve of IQ,which includes but is not limited to micropheniceffects. In fact, refinements in diagnostic tech-nique may lead to recognition of megaphenic causesunderlying an ostensibly Gaussian distribution, likethe three alleles which account for most of thevariation of red cell acid phosphatase (Harris, 1975).Segregation analysis of mental defect can do little
3
with the microphenic component except determineits frequency and recurrence risks, leaving cumu-
lative environmental and genetic factors to be re-
solved by path analysis in the general population(Rao et al, 1974, 1975). Here we shall not attemptto partition this microphenic component, and esti-mates of heritability will, therefore, include culturalas well as biological inheritance. The goal of theanalysis is to delimit the microphenic contributionand isolate components of the megaphenic categorycaused by genetic mechanisms.To that end an index of severity is useful, since
the proportion of megaphenic cases increases withseverity. We used the method of criterion scaling(Beaton, 1969), recording grade of patient from 100for dull to 400 for idiot and transforming all othervariables in terms of the mean value of this cri-terion (Table III). Variables used by Penrose butnot tabulated in his appendix were extracted fromthe original records by one of us (H.L.B.). Con-sanguinity was not significantly related to severityand so was omitted. The 14 significant variableswere combined by the first principal component oftheir correlation matrix.
If the assumed categories are valid, the distribu-tion of the severity index should be a mixture ofmegaphenic and microphenic components. Mac-Lean et al (1976) have developed an analysis of ad-mixed distributions which allows for skewness. Itwas used to define a most sociofamilial group (BO)corresponding to 9%0 of cases. This is consider-ably less than the category without significantphysical or metabolic pathology (24%), which mustcontain megaphenic cases (aminoacidopathies, etc.)not recognized with the diagnostic techniquesavailable at the time of the Colchester study, and in
TABLE IIICRITERION SCALING OF PROBANDS: UNKNOWN AND OTHER DENOTED BY *
Variable i Scale (zi) ai bi
Sex 1 &-262, Y,245 8.5 0.076Grade 2 MO-100, M1-÷200, M2-+300, M3-+400 93.4 0.267Reliability 3 1-*261, 2-*236, 3--238 11.0 0.297Grade of father 4 S--302, N-*261, MO-+236, M1-.227, M2-*233, *+208 15.1 0.352Grade of mother 5 S--332, N-+265, MO-+230, M1-+229, M2--261, *+217 16.9 0.328Age 6 394- 13.5A + 0-35A2 - 0.0025A3 22.0 0.113Admission 7 1--308, 2-*308, 3+206, 4-*263, 5-+270, 6-s183 38.7 0.224Occupation of father 8 A1-+306, A2-÷284, A3-s298, B-277, C-*277, D- 290, El-+256, E2-.244, *+222 19.3 0.326Age of father 9 230 + 0.83A, *-+223 12.4 0.313Age of mother 10 228 + 0.0285A2, *-+229 12.9 0.215Rating of home 11 A-*328, B-293, C-÷257, D-*237, E-+222, *--*242 20.8 0.332Size of home 12 233+ 7.54R-2.94P, *--->248 10.7 0.236Illegitimacy 13 1-+222, 0-257 9.3 0.281AAMD diagnosis 14 Olx- 276, 02x--219, lxx-+276, 2xx-*275, 3xx--259, 4xx-*309, 46x-s240,
The equation A-N signifies that probands in category A are assigned their mean severity N. For example males are scored as 262. Theseverity index is computed as Y = jbjzj/aj, where ao is the standard deviation and b is the eigenvector (i = 1,...,14).
many cases probably not recognizable at the pre-sent time.We also created an index of biological factors,
taking as the criterion the value 1 for AAMD codesbeginning with 2-7 or 9, considered to be mostlybiological, and the value 0 for other AAMD codes(0, 1, 8) with large medical or sociofamilial com-ponents. For this purpose the Colchester datawere reclassified by the AAMD code (Grossmanet al, 1973) as used by us for mental retardation inHawaii (Table IV). The resultant biological indexhas so high a correlation with the severity index(0.963) that it is impractical to separate biologicaland medical cases on the basis ofvariables other thanmedical history.The estimate ofadmixture for the biological index
is 0.84, leaving 16% as medical and most socio-familial. Since the latter were estimated from theseverity index to comprise 9% of cases (115 indi-viduals), we conclude that about 90 cases (7%)should be classified as medical. Accordingly wetook as the medical category (M) the 90 largestvalues of the severity index for diagnostic codesbeginning with 0 or 1. In the remainder the
115 smallest values of the severity index were takenas most sociofamilial (Bo) and the rest (84%) weredivided into quartiles from least to most mega-phenic (B1 to B4).The most megaphenic biological group (B4) con-
tains a substantial number of patients classified byPenrose as without physical or metabolic pathology(Table IV). Most diagnoses contributed fewmembers to Bo, which is drawn largely fromaclinical cases, mild psychopathy, and congenitalsyphilis. The latter was diagnosed by a variety ofcriteria, and the possibility of false positives wasdiscussed by Penrose (1938, pp. 38-40), who con-cluded that 'in some cases defect may be entirelydue to the physical disease but in others it merelyaccentuates an endogenous mental inferiority'.Our analysis, in which 16 of 50 cases are classifiedas most sociofamilial (BO), supports this suggestion.There is a clear relation between aetiological
group and severity of retardation, which increasesfrom B1 to B4 (Table V). The proportion ofsociofamilial cases decreases from mild to severedefect. Most cases classified as extremely socio-familial (BO) have one or both parents of unclassified
TABLE VGRADE BY AETIOLOGY
Grade |IQ Range Medical I_|Sociofaniial Mixed Biological TotalBo B, B2 B, B Toa
intelligence (Table VI). They are often illegiti-mate and with few if any full sibs. The proportionof cases with one or both parents dull or retarded ishigh for B1 and decreases to virtually zero for B4,the most biological group.However interpreted, the aetiological categories
are more homogeneous than the whole sample.Segregation and consanguinity analysis can tell usmore about them.
Segregation analysis of aetiological groupsTable II gives 0.0153 as the frequency of re-tardation in the general population if incidence isappropriate for severe defect, and prevalence formild defect. We consider a patient of grade Mo(dull) or worse and a parent or nonpatient sib ofgrade M1 (mildly retarded) or worse as affected.With this definition the ascertainment probabilityfrom the distribution of the number ofprobands persibship in the 222 different sibships with two or moreaffected is 7T= 0.263 ± 0.029 (Morton, 1959). Thisestimate does not vary significantly among aetio-logical groups, presumably because mortality ofsevere cases and community placement of mild caseshave somewhat similar effects on the ascertainmentprobability. Therefore, we assume the above valueof vT for each category. For lack of evidence aboutthe frequency in different segments ofthe populationwe also assume the same incidence for each cate-gory.A preliminary analysis of severe mental defect by
complex segregation analysis (Morton and Mac-Lean, 1974) agreed closely with the simpler analysisof Dewey et al (1965). It also showed that par-titioning individuals into 3 classes (normal, inter-mediate, affected) gives greater power than twoclasses (normal, affected). Since we are now con-sidering all grades of mental retardation, we defineintermediate by the frequency of dull but not re-tarded individuals in the general population. TableI gives 423 dull individuals among 10 888 second
degree and more remote relatives, a proportion of0.0388. Classification on the psychometrics ratherthan conservative judgment might raise this toapproximately the frequency in a normal distribu-tion between 1 and 2 standard deviations (IQ 85-70),or 0.136. A lower frequency is appropriate for theColchester Survey, where the diagnosed frequencyof dull sibs was only 0.106 (cp. Reed and Reed,1965, p. 5). For the displacement of recessivehomozygotes we take t = 3.5. This is supported byanalysis of normal x normal matings which give aslightly smaller likelihood at lower values of homo-zygous displacement t when heritability H and re-cessive gene frequency q are estimated simultane-ously. At t= 3.5 the other estimates are H = 0.30 ±0.03, q = 0.03 ± 0.01.
Table VII gives the analysis of aetiological cate-gories, separating congenital syphilis and Down'ssyndrome from the rest of the material and poolingthe most sociofamilial categories Bo, B1, B2, whichon preliminary analysis appeared to be homo-geneous in regard to the parameters for heritabilityH and recessive gene frequency q. This socio-familial group gives no evidence of recessive genes,but a nonsignificant suggestion of environmentcommon to sibs. Cultural inheritance from parentto child is confounded with heritability, the estimateof which is 0.833 ± 0.026. While the hypothesesthat heritability in the sociofamilial group is en-tirely genetic or entirely cultural cannot be dis-proven on these data, such limiting cases seem un-likely since IQ is determined by both genetic andcultural inheritance (Rao et al, 1974).The most megaphenic group (B4) gives highly
significant evidence for recessive genes (q= 0.048 +0.009), but no suggestion ofheterozygous expressionor polygenic heritability. Since there is no signifi-cant role for other familial mechanisms, most non-recessive cases are sporadic, with negligible recur-rence risks.The intermediate group (B3) seems to be a mix-
Frequency of affected (A) = 0.0153, frequency of intermediates (I) = 0.0388, dominance (d) =0, displacement (t) =3.5,and ascertainment probability (ir) = 0.263. The columns indicated as x2 are computed as - 2lnL + const. For example,the test of q=0 in group B4 is x' = 392.77 -369.71 = 23.06.
intermediate heritability (0.603 ± 0.061). There isno significant evidence of recessive genes, but thestandard error of the gene frequency is large. Ifweassume admixture in proportion x of families withmegaphenic cases indistinguishable from B4 (H= 0,q= 0.048), the remaining fraction 1- x of familieshaving microphenic cases (H= 0.833, q= 0), the
likelihood is appreciably increased (X2 = 970.55).Genetic counselling in this category should, there-fore, be made on the hypothesis of admixture ratherthan from the intermediate values of H and q esti-mated directly. The maximum likelihood estimateof the frequency of megaphenic cases in B3 is0.59±0.09.
TABLE VIIIPARENTAL CONSANGUINITY BY ETIOLOGY
Inbreeding Coefficient F Toal Mean FAetiology Tota x l1050 1/128 1/64 1/32 1/16 1/8 1/4
Colchester revisited: A genetic study of mental defect
InbreedingTable VIII gives clear evidence that inbreeding,and especially incest, is more frequent for dull orretarded individuals than for normals. Consan-guinity is conspicuously raised among the parents ofthe sociofamilial group, but is not appreciably morefrequent than in the control population when onlynormal x normal matings are considered.
For B3 + B4, on the other hand, the strikingassociation with parental inbreeding persists innormal x normal matings. If we take 0.0128 as theincidence in these matings, the method of Morton(1960) gives
A=panmictic load=0.0124
B = inbred load = 0.893 + 0.263
For Swedish data (Book, 1957) the correspondinginbred load is
B = 0.716 ± 0.349
and so the pooled estimate is B = 0.83 ± 0.21.Since this is about four times the load for severemental defect, it confirms the suggestion of Deweyet al (1965) that 'some rare recessive genes act toproduce mild retardation, though they are difficultto discriminate because of overlap with the poly-genes and environmental factors causing the lowerpart of the continuous, nearly normal distribution ofintelligence'.The incidence of homozygotes in the general
population served by Colchester has been estimatedas 0.0482 (Table VII). Deducting Ba, the ex-
pressed load resulting from an inbreeding coefficientca, we have
A=1964x 10-6as the panmictic incidence of mental retardationcaused by rare autosomal recessive genes, or sixtimes the estimate of Dewey et al (1965, p. 254) forsevere mental defect. The mean gene frequencyper contributory locus is Q < A/B = 0.0024, and thenumber of contributory loci is k. B2/A = 351.These estimates could easily be in error by a factorof 2, but they indicate that the number of loci islarge and the individual gene frequencies small.Assuming a mutation-selection equilibrium, themutation rate per gamete is about 0.O1B = 0.008 andthe mutation rate per locus is ,u < 0.O1Q = 2.4 x 10 5,in reasonable agreement with other rare recessive
genes. The possibility that heterozygote advantageplays an important role in maintaining this load isremote (Dewey et al, 1965). In that event, themutation rates would be smaller but the estimates ofgene frequency and number of loci would not beaffected.
7
Sex ratioIf sex linkage accounts for part of the excess ofretarded males, as has been suggested by Lehrke(1974), there should be an excess of families withonly affected males corresponding to the distribu-tion
where h is the proportion unable to segregateaffected girls, p is the probability of girls amongaffected, r is the number of affected girls, and s - r isthe number of affected boys (Morton, 1959). Forthe 1227 Colchester families there is no doubt thatmales are in excess among affected (XI = 14.6), butthe proportion of cases caused by sex linkage is indoubt (Table IX). The maximum estimate of theproportion of sex-linked families h for p < 0.5 is atthe limit p = 0.5, with h = 0.089 ± 0.023, and thelikelihood is virtually the same as for h and p esti-mated simultaneously. The hypothesis that thesex ratio p = 0.5 is admissible, and among normalsibs the proportion of females is 0.494 ± 0.008, as inthe general population. However, it is unlikely thatboys and girls are equally frequent among cases notdue to major sex-linked genes, since retarded boysare more difficult to manage and polygenic vari-ability must be greater for the heteromorphic sex.If polygenes are distributed between autosomes andthe X chromosome in proportion to their length andif the heritability of IQ in the general population is0.5 (Rao et al, 1974), then the sex ratio of affectedwould be
0.494P= 0.494 + 0.506eo.025 = 0.488,
at which value the estimate of h is 0.070 ± 0.024.This is not significantly different from the estimateof 0.040 + 0.046 when p and h are iterated simul-taneously. We tentatively conclude that about 5%of cases are the result of sex linkage.
DiscussionThis analysis depends on the basic soundness ofthe Colchester survey, which includes data essential
TABLE IXTHE SEX OF AFFECTED
Hypothesis x2
Equal sex ratio, no sex-linkage (p = 0.5, h = 0) 37.13Equal sex ratio, sex-linkage (p = 0.5, h = 0.089 ± 0.023) 23.20Unequal sex ratio, no sex-linkage (P = 0.452 ± 0.013, h 0) 22.52Unequal sex ratio, sex-linkage
(fi = 0.470 ± 0.025, f = 0.040 ± 0.046) 21.95Theoretical sex ratio, sex-linkage
for resolution of heterogeneity and complex segre-
gation analysis. However, with the wisdom of 40years' hindsight some refinements can be suggested.Replacement of intellectual grades by psycho-metrics on all probands, sibs, and parents wouldallow more powerful segregation analysis and more
reliable estimates of incidence. Greater sympto-
matic detail (and of course application of the cyto-genetic, biochemical, and other diagnostic advancesof the last generation) would give better resolutionof aetiological heterogeneity.These data provide no critical distinction between
cultural inheritance and polygenes in the large pro-
portion of sociofamilial cases. The distinction,while interesting, is irrelevant to recurrence risks infirst-degree relatives, which can be predicted fromthe values of segregation analysis (Maclean et al,1976).On the evidence there is clear separation of the
extreme groups of sociofamilial retardation (Bo+B1+B2) from megaphenic defect (B4) after exclu-sion of recognized exogenous types (M). Thedistribution of cases is estimated in Table X for theColchester survey and the general population.The majority are sociofamilial, only about 7% are
the result of recognized exogenous factors, and per-
haps a third are biological. Since premature mor-
tality reduces prevalence of severe defect by aboutone-half and incomplete diagnosis has roughly thesame effect on prevalence of mild defect, the pro-
portions are much the same for incidence and pre-
valence. The medical category would presumablybe more frequent under the modern convention ofclassifying moribund young children with multiple
anomalies as severely retarded, whereas they were
excluded in the Colchester survey.
The incidence of homozygotes for rare autosomalrecessive genes causing mental retardation was esti-mated above as 0.0482 = 0.0023. The incidence ofbiological cases (B4 + 0.59 B3) is 0.0052. Apparentlyautosomal recessives account for less than half of thebiological category, the remainder being due to othermechanisms such as sex-linkage, recent dominantmutations (both genic and chromosomal), aneu-
ploidy, unrecognized medical factors, and perhapsmore complicated causes.With respect to sex-linkage our results are incon-
clusive, since the distribution of affected males andfemales is consistent with a negligible or a consider-able proportion of sex-linked cases, depending on
the sex-ratio of affected in other cases (p). If 5% ofgenetic cases are the result of sex-linkage, as mightbe expected if the loci which produce mental re-
tardation are randomly distributed among chromo-somes (Paris Conference, 1971), then the incidenceof sex-linked recessive types of mental defect wouldbe less than 0.0003. However, this neglects thereduction in incidence of homozygotes for autoso-mal recessives as inbreeding declined during the lastcentury. If the mean gene frequency is only 0.0024per contributing locus (as estimated), then mosthomozygotes at a mutation-selection equilibriumwere due to identity by descent. If so, at presentreduced levels of inbreeding sex-linked recessivesmight make up several per cent of all cases, withoutany disproportionate concentration on the X chro-mosome of genes for mental retardation. At 5%of all cases the number of Colchester probands
TABLE XESTIMATED DISTRIBUTION OF CASES AND INCIDENCE
Colchester revisited: A genetic study of mental defect
would be 64. If these fall into the B3 and B4 cate-gories they might from Table V have an incidence of2(64)(0.00724)/534, or 174 per 100 000 male births.If fertility of these hemizygotes is negligible, butheterozygous females are normal, and if mutation isequally frequent in egg and sperm, the mutationrate per X chromosome would be 58 x 10-5. Weestimate the number of contributory loci per auto-somal genome to be at least 351. If the X chromo-some carries a proportionate number of recessivegenes there would be at least 18 X-linked loci thatcould cause mental defect and the mutation rate perlocus would be at most 3.2 x 10-5 a typical value.
Although any attempt to partition the biologicalcategory must be approximate, it seems worthwhile to make the attempt (Table XI). We havejust estimated the sex-linked component and theincidence of autosomal recessives. If the numbersof cases are proportional to the incidences, as wouldbe expected if sex-linked and autosomal recessiveshad on average similar degrees of severity, therewould be about 169 autosomal recessive cases in theColchester survey, of whom 22 have normal butconsanguineous parents (Table VIII).Among the coded diagnoses are 6 cases of neuro-
fibromatosis, 5 of epiloia, 1 of Huntington's chorea,and 21 of acrocephaly (including acrocephalo-syndactyly). From Bergsma (1974) the incidenceof dominant types of mental retardation is about 40per 100 000. On the same assumptions as for re-cessives, this would give about 30 dominant cases inthe Colchester survey.Newton et al (1972) conducted a cytogenetic sur-
vey of 1255 patients over 15 years of age in a hospitalfor the mentally subnormal. There were 104 casesof Down's syndrome, in rough agreement with the72 cases of the Colchester survey in samples ofvirtually the same size. The incidence of Down'ssyndrome is about 138 per 100 000 births (Penrose,1963, p. 208). There were 24 other chromosomalabnormalities in the cytogenetic survey, comparedwith 422 per 100 000 in the general population(Jacobs et al, 1974), of which about 1/10 causemental retardation (P. A. Jacobs, 1976, personalcommunication).Adding the genetic categories together, we ob-
serve that their frequency is approximately B4 +0.59 B3, as expected. Most of the biological groupis the result of simple mechanisms, with little if anyresiduum where unrecognized exogenous or complexcauses could be involved.
9
REFERENCES
Beaton, A. E. (1969). Scaling criterion of questionnaire items.Socio-Economic Planning Sciences, 2, 355-362.
Bell, J. (1940). A determination of the consanguinity rate in thegeneral hospital population in England and Wales. Annals ofEugenics, 10, 370-391.
Bergsma, D. (1974). Birth Defects: Atlas andCompendium. Williamsand Wilkins, Baltimore (The National Foundation).
Book, J. A. (1957). Genetical investigations in a north Swedishpopulation: the offspring of first-cousin marriages. Annals ofHuman Genetics, 21, 191-221.
Crow, J. F. (1968). Some analysis of hidden variability in Droso-philia populations. In Population Biology and Evolution, pp. 71-86. Ed. by R. C. Lewontin. Syracuse University Press, NewYork.
Dewey, W. J., Barrai, I., Morton, N. E., and Mi, M. P. (1965).Recessive genes in severe mental defect. American Journal ofHuman Genetics, 17, 237-256.
Grossman, H. J., Warren, S. A., Begab, M. J., Eyman, R., Nihira, K.,and O'Connor, G. (1973). Manual on Terminology and Classi-fication in Mental Retardation (Revised). American Association ofMental Deficiency, Special Publications Series 2.
Harris, H. (1975). The Principles of Human Biochemical Genetics.Elsevier, New York.
Jacobs, P. A., Melville, M., Ratcliffe, S., Keay, A. J., and Syme, J.(1974). A cytogenetic survey of 11,680 newborn infants. Annalsof Human Genetics, 37, 359-376.
Lehrke, R. G. (1974). X-linked Mental Retardation and VerbalDisability. Birth Defects: Original Article Series, 10, No. 1.Intercontinental Medical Book Corporation, New York.
MacLean, C. J., Morton, N. E., Elston, R. C., and Yee, S. (1976).Skewness in commingled distributions. Biometrics. In the press.
Morton, N. E. (1959). Genetic tests under incomplete ascertain-ment. AmericanJournal ofHuman Genetics, 11, 1-16.
Morton, N. E. (1960). The mutational load due to detrimentalgenes in man. American Journal of Human Genetics, 12, 348-364.
Morton, N. E. (1974). Analysis of family resemblance. I. Intro-duction. AmericanJournal ofHuman Genetics, 26, 318-330.
Morton, N. E. and MacLean, C. J. (1974). Analysis of family re-semblance. III. Complex segregation of quantitative traits.American Journal ofHuman Genetics, 26, 489-503.
Newton, M. S., Cunningham, C., Jacobs, P. A., Price, W. H., andFraser, I. A. (1972). Chromosome survey of a hospital for thementally subnormal. Part 2: Autosome abnormalities. ClinicalGenetics, 3, 226-248.
Paris Conference (1971). Standardization in human cytogenetics.Birth Defects: Original Article Series, 8, No. 7. The NationalFoundation, New York, 1972.
Penrose, L. S. (1938). A clinical and genetic study of 1,280 cases ofmental defect. Medical Research Council. Special ReportSeries, No. 229.
Penrose, L. S. (1963). The Biology of Mental Defect. Sidgwickand Jackson, London.
Rao, D. C., MacLean, C. J., Morton, N. E., and Yee, S. (1975).Analysis of family resemblance. V. Height and weight in north-eastern Brazil. American3Journal ofHuman Genetics, 27,509-520.
Rao, D. C., Morton, N. E., and Yee, S. (1974). Analysis of familyresemblance. II. A linear model for familial correlation. Ameri-can3Journal ofHuman Genetics, 26, 331-359.
Reed, E. W. and Reed, S. L. (1965). Mental Retardation: A FamilyStudy. W. B. Saunders, Philadelphia.
Tarjan, G. (1970). Some thoughts on socio-cultural retardation.In Socio-cultural Aspects of Mental Retardation. Ed. by H. C.Haywood. Appleton-Century-Crafts, New York.
Tarjan, G., Wright, S. W., Eyman, R. K., and Keeran, C. V. (1973).Natural history of mental retardation: some aspects of epidemio-logy. American_Journal of Mental Deficiency, 77, 369-379.