146 Correspondence Table 1.

Before After Normal (n = 2 5) IFN therapy JFN therapy

PB neutrophil phenotype My9 (CD33) 11.4f 2.4% 28.8% 15.3% Mol (CDl1) 84*6f2*9% 32.6% 65.1% Leu 15 (CDl1) > 90% 60.2% 67.3%

NK cells (Leu Ila+) % PBM 16.7 f 2% 26% 38% Absolute count 0*5f0*1 x 109/1 0.37~ 1O9/I 0.45 x lo9/]

NK activity at various effector:target ratios 50: 1 46.6f6.2% 12.9% 34% 25:l 40.3 f 6.3% 9% 24.7% 15:l 32.5f 5.2% 6.2% 18.2% 10: 1 24f4.0% 5% 10.3% 5:l 15.3f2% 4.8% 6.6%

11 6 x 109/1. Serum B12 and folate were normal. Bone marrow revealed dyserythropoiesis (but no ring sideroblasts) and dysmyelopoiesis with Pelger-Huet cells and the following myelogram: neutrophils 18%, lymphocytes 7%, nucleated red cells 35%, monocytes 2%, myelocytes 5%. metarnyelo- cytes 2%, plasma cells 2%. promyelocytes 1%. blasts 18%. A diagnosis of refractory anaemia with an excess of blasts (RAEE3) was made, with the poor prognostic features (anae- mia, neutropenia, thrombocytopenia. excess blasts) defined by Mufti et al(1985). The patient was transfused with 5 units of blood and commenced on 3 megaunits of a-IFN (Wellferon) subcutaneously each day.

Following an initial fall in neutrophil count this rose to 2-2 x 109/1 by 6 weeks and 4.6 x 10y/l by 4 months therapy. His haemoglobin followed a similar pattern stabilizing at around 12 g/dl. Platelets remain around the 80 x 109/1 mark. Bone marrow examination at 6 weeks revealed less dyseryth- ropoiesis, and myelopoiesis was improved (neutrophils 4 1 %, lymphocytes lo%, nucleated red cells 17%, monocytes 5%, myelocytes lo%, metamyelocytes 8%, plasma cells 4%, promyelocytes 2%. basophils 0%. blasts 2%).

Indirect immunofluorescence revealed increased periph- eral blood (PB) neutrophil maturity (Baumann et al. 1986) and natural killer (NK) cells following 12 weeks Wellferon therapy (see Table I). NK function, as determined by a standard 51chromium release assay, was also improved after Wellferon therapy.

The initial side effects of Wellferon were well tolerated and resolved quickly.

Universitu Department of Haematology, Duncan Building. Roual Liverpool Hospital. Prescot Street. PO Box 147, Liverpool L69 3BX

Wellcome Research Laboratories, Lungleu Court. Beckenham, Kent

D. W. GALVANI J. C. CAWLEY

A. NETHERSELL J. M. BOTTOMLEY

REFERENCES

Baumann, M.A., Keller, R., McFadden, P.W., Libnock, J. & Patrick, C.W. (1986) Myeloid cell surface phenotype in myelodysplastic states: evidence of an early diEferentiation antigen. American j o u m l ojHernuto1ogM. 22,251-257.

Mufti. G.J., Stevens, J.R.. Oscier, D.G., Hamblln. T.J. & Machin, D. (1 98 5) Myelodysplastic states - a scoring system with prognostic significance. British j o u m l ofHaernatolog#. 59,425433.

Okabe, K.. Minagawa, T.. Nakane. A., Sakurada, K. & M i y d i . T. (1986) Impaired ainterferon production and natural killer activity in blood mononuclear cells in myelodysplastic syndromes. Scandf- navfan journal oj Huernatologll, 37. 1 11-1 17.

BAYESIAN ANALYSIS OF DNA HAPLOTYPE ASSOCIATIONS IN p THALASSAEMIA AND APPLICATIONS TO PRENATAL DIAGNOSIS

The article by Wainscoat et a1 (1 986) on prenatal diagnosis of /3 thalassaemia demonstrates again how genetic linkage can be applied to prenatal or presymptomatic diagnosis of diseases. DNA polymorphisms occur at various restriction enzyme sites in the B globin gene cluster (Kan 81 Dozy, 19781,

and together these identieable polymorphisms form a marker haplotype. Once a particular marker haplotype is found to be Linked to a thalassaemic allele in a family, the marker haplotype can be used to prenatally identify thalassaemic alleles transmitted within that family. The article also

illustrates the concept of linkage disequilibrium, or allelic association, between a disease allele and specific chromoso- mal haplotypes (Antonarakis et al, 1982). Linkage disequili- brium involving thalassaemia presumably reflects the posi- tive selection, due to malaria, for thalassaemia heterozygotes and the consequent enrichment in the population for the chromosomal DNA sequences upon which the thalassaemic mutations originally arose (Kazazian et al, 1984). The result is that the thalassaemia alleles in the population are found preferentially on number 1 1 chromosomes with specific marker alleles or haplotypes.

The linked and associated DNA markers for /3 thalassaemia have parallels to the HLA markers for hereditary haemo- chromatosis. Hereditary haemochromatosis is a Mandelian recessive disease (Saddi & Feingold. 1974) characterized by parenchymal iron overload in a variety of tissues. As many as 5-10% of individuals in the general population are heterozy- gotes for the haemochromatosis gene (Beaumont et al, 1979). The gene is tightly linked to the HLA region on chromosome 6 (Simon et al, 1977: Kravitz et al. 1979), so that HLA- identical siblings of an affected individual are considered to be homozygous for the gene and presymptomatically affected. The haemochromatosis gene is furthermore preferentially associated with particular HLA haplotypes in the population (e.g. A3,B14: Simon et al. 1976). Because of the high prevalence of carriers, the use of the preferential associations in conjunction with the family linkage method has been proposed as an approach to assessing risks for homozygosity for relatives beyond HLA-identical siblings before conven- tional iron studies are diagnostic (Lin er al, 1985).

When complete linkage and association information are available for /3 thalassaemia, it should be similarly desirable to calculate the risks for thalassaemic alleles for various DNA haplotypes from population association data. For example. if both parents are thalassaemia heterozygotes, what is the risk that either of them has transmitted a thalassaemic allele to their first offspring? Or. what is the risk that the offspring will be affected? The risks are 50% and 25%, respectively, if only Mendel's rules are considered. Importantly, without a prior affected child, the existence of linkage with marker haplo- types is of no utility. However. these figures can be consider- ably refined by taking into account the DNA haplotype population associations.

The probability that a chromosome in the general popula- tion carries a thalassaemic allele when it carries a particular haplotype is given by Bayes' theorem (Feller, 1968):

P(B""IH) = p ( P ' ) x P(HIP")/P(H) where p(bthal) is the gene frequency for /3 thalassaemia or, in probabilistic terms, the prior probability for a thalassaemic allele. The term, P(HI/~*~'). stands for the frequency with which a specific haplotype is found on a chromosome that carries a thalassaemic allele, and p(H) is the frequency of the designated haplotype among all chromosomes in the general population. In principle then, the clinically-important proba- bility for a thalassaemic allele given a specific haplotype (p(PllH), known as the posterior probability), can be calculated by substituting the results provided by Table I of Wainscoat et al into the Bayes' theorem formula. We illustrate this application here. The haplotypes

Correspondence 14 7 +-- - [ - I -++ and - + + - [ + I + + + are high risk haplotypes, which we shall consider as a group because of the relatively small sample sizes if data are considered for each haplotype separately. The high risk haplotype group has

p(HIfltha1) = (1 5 + 22)/50 = 0.74

The overall probability of the haplotype group (designated by H) is approximately the frequency of the haplotypes among the control, nonthalassaemic individuals, so that

p(H)--p(HI/3')= (0+ 1)/50=0.02

If the gene frequency for /3 thalassaemia in Milan is 0.01 (Livingstone, 1985). then

p(~"lH)=O.Ol x 0.74/0.02=0*37

That is, a random chromosome in the population carrying either the +- - - [ - I -++ or the - + + - [ + I + + + haplotype has a 37% probability of having a thalassae- mic allele. The haplotypes + - - - [ +] - + + , +-- [+I -+- and - + - + [ + I + + + are low risk- haplotypes. Similar calculations with the combined data for these latter haplotypes yield a 0.2% probability for a thalassaemic allele on a random chromosome carrying any of these low risk haplotypes. Population association probabili- ties such as these can be obtained for any haplotype for which population data are available.

In order to extent use of population association probabili- ties to risk estimation for the fist offspring of parents who are known to be heterozygotes for B thalassaemia. the population association probabilities must be normalized to account for the fact that each parent has exactly one thalassaemic chromosome. To normalize, one must know the haplotypes for each of the parental chromosomes and use them to determine each chromosome's relative risk for carrying that parent's thalassaemic allele. The relative risks for carrying the thalassaemic allele must sum to one for each parent. Hence, if a heterozygote parent has haplotypes H1 and H2, the risk for a thalassaemic allele on the chromosome that bears haplotype H1 is

PM"" I H1 )/UPll H 1 1 +p(Bhal I H2)I and the risk for a thalassaemic allele on the other chromo- some, which bears haplotype H2. is

p(P""l H2)/b(BhallH1 1 +p(P"lH2)1

For example, if a heterozygote parent has one chromosome with a high risk haplotype. e.g. + - - -[ - J - + +, and one chromosome with a low risk haplotype, e.g. + - - - [ +] - + +, and transmits the chromosome with the high risk haplotype, then the risk to the offspring for inheriting a thalassaemic allele from that parent is approxi- mately 0*37/(0-37 +0.002), or 99%. If the offspring receives such high risk hapIotypes from both parents, the risk for thalassaemia would be (0-99)2, or 98% rather than the prior Mendelian risk of 25%. On the other hand, if the parent transmits the chromosome with the low risk haplotype, then the risk for a thalassaemic allele is 0.002/(0.37 + 0.002), or 1%. These estimates can thus yield dramatic refinements of the Mendelian risk. Inspection of the formulas for the chromosomal risks show that when a parent's two haplo-

148 Correspondence types are identical, then the calculations simply reproduce the Mendelian risk of 50% for transmissfon of a single thalassaemic allele.

A few comments should be made regarding our application of the Bayesian method. First, the parents in the families reported by Wainscoat et al may not reflect a cross section of Milan, the catchment area under study. Some of the parents had diverse origins outside of Milan, and the families appear not to have been randomly chosen from all those at risk for a second child with fl thalassaemia. Second, the reported data are probably not extensive enough to produce statistics usable for diagnoses. Standard errors of the calculated risks are apt to be large. This can be remedied by increasing the sample size. And third, the gene frequency for fl thalassaemia varies widely from one geographic area to another, and the frequencies with which speciclc haplotypes are associated with thalassaemia also show geographic clustering (Orkin et al, 1982). The exact figures to be used in the calculations should therefore take into account the geographic origins of the parents. To improve the calculations, one needs large compilations of haplotype frequencies among various control populations as well as among thalassaemic individuals.

Because of the interest in fl globin cluster haplotypes and prenatal diagnosis of thalassaemia, statistically reliable h a p lotype frequencies should become available if each centre records and periodically analyses its cumulative data. Such data can then be used to construct tables that will be of practical value for risk assessment and prenatal diagnosis following heterozygote screening. The approach will also be useful for assessment of risks to other classes of relatives of index cases, e.g. the offspring of an aunt or uncle of an index case. Currently, the calculations based on Bayes’ theorem quantitate the strengths of DNA haplotype associations in fl thalassaemia, indicating which DNA haplotypes are likely to be informative and which not for prenatal detection. Analysis of more extensive haplotype data will hopefully allow these Bayesian methods to be increasingly useful in the diagnostic setting.

Clinical Hernatologg Branch, HENRY J. h N National Heart. Lung, and Blood Institute, National Institutes of Health. Building 10 Room 7C103 Bethesdu, MD 20892, U.S.A.

Departments of Medicine and Pediatrics, UCLA School of Medicine. Division of Medical Genetics, Cedars-Sinai Medical Center, 8700 Beverleu Blvd., Los Angehs, CA 90048, U.S.A.

JEROME I. R m

Announcement Festchrift for Professor David A. G. Galton. MD, FRCP, OBE

A oneday Scientific meeting on Acute and Chronic Leukae- mias followed by a dinner in honour of Professor D. A. G. Galton, Director of the MRC Leukaemia Unit, on the occasion of his retirement, will take place at the Royal Postgraduate

REFERENCES

Antonarakis, S.E., Boehm. C.D., Giardina. P.J.V. & Kazadan, H.H.. Jr (1 982) Nonrandom association of polymorphic restriction sites in the /?-globin gene cluster. Proceedings of the National Academy of Sciences of the United States of America. 79, 137-141.

Beaumont, C., Simon, M., Fauchet, R., Hespel. J-P., Brissot, P.. Genetet, B. & Bourel, M. (1979) Serum ferritin as a possible marker of the hemochromatosis allele. New England Journal of Medicine,

Feller. W. (1968) An Introduction to Probabilit~ Theory and its Applications. Vol. 1, 3rd edn. John Wlley and Sons, New York.

Kan. Y.W. & Dozy. A.M. (1978) Polymorphism of DNA sequences adjacent to human /%globin structural gene: relationship to sickle mutation. Proceedings of the National Academy of Sciences of the United States of America, 75, 5631-5635.

Kazazian, H.H., Jr, O r h , S.H., Markham, A.F., Chapman, C.R., Youssouflan, H. & Waber. P.G. (1984) Quantification of the close association between DNA haplotypes and speciec fi-thalassaemia mutations in Mediterraneans. Nature. 310, 152-1 54.

Kravitz, K., Skolnick, M.. Cannings, C., Carmelli, D.. Baty, B.. Amos, B., Johnson, A.. Mendell. N., Edwards, C. & Cartwright, G. (1979) Genetic linkage between hereditary hemochromatosis and HLA. American Iournal oJHuman Genetics, 31, 601-619.

Lin, H.J.,Conte, W.J.&Rotter, J.I.(1985)Mseaserfsksestimatesfrom marker association data. Application to individuals at risk for hemochromatosis. Clinical Genetics, 27, 127-1 3 3.

Livtngstone. F.B. (1985) Frequencies of Haemoglobin Variants. Oxford university Press.

OrkIn. S.H., Kazadan. H.H., Jr, Antonarakis, S.E., Goff. S.C.. Boehm. C.D..Sexton. J.P., Waber,P.G.&Giardina.P.J.V. (1982)Linkageof /?-thalassaemia mutations end /%globin gene polymorphisms with DNA poiymorphisms in human /?-globin gene cluster. Nature. 296,

Saddi. R. & Feingold, J. (1974) Idiopathic haemochromatosis: an autosomal recessive disease. Clinical Genetics, 5, 234-241.

Simon, M.. Bourel. M.. Fauchet, R. & Genetet. B. (1976) Association of HLA-A3 and HLA-Bl4 antigens with idiopathic haemochroma-

Simon, M., Bourel, M. Genetet, B. & Fauchet, R. (1977) Idiopathic hemochromatosis. demonstration of recessive transmission and early detection by f d y HLA typing. New England Journal of Medicfne, 297, 1017-1021.

Wainscoat. J.S., Work, S.. Sampietro, M., Cappellini. M.D.. Fiorelli, G., Temli. S. & Weatherall. D.J. (1986) Feasibility of prenatal diagnosis of /? thalassaemia by DNA polymorphism in an Italian population. British Journal of Haematdog#, 62, 495-500.

301,169-174.

62 7-63 1.

h i s . Gut. 17, 332-334.

Medical School, London. W.12, on Friday, 9 October 1987. People interested in participating should contact Dr D. Catovsky, MRC Leukaemia Unit, Royal Postgraduate Medical School. Ducane Road, London W12 OHS. Telephone: 740 3237 (direct h e ) .