M. Hickman, CRDHB, Imperial College School of Medicine, 200 Seagrave Road. London SW6 1RQ.
A range of estimates for sickle cell and β thalassaemia have been derived for the different ethnic groups living in the U.K., reflecting uncertainty over the true population value in certain countries and the heterogeneity within and between countries of origin comprising the same ethnic group. These were validated against six community screening programmes, with the estimated range correctly predicting the number of affected births observed by the programmes.
In England approximately 3000 affected babies (0.47%) carry sickle cell trait and 2800 (0.44%) carry β thalassaemia trait annually; with approximately 178 (0.28 per 1000 conceptions) affected by sickle cell disease (SCD) and 43 (0.07 per 1000) by β thalassaemia major/intermedia. Allowing for termination, about 140–175 (0.22–0.28 per 1000) affected infants are born annually with SCD and from 10 to 25 (0.02–0.04 per 1000) with β thalassaemia major/intermedia.
These are the first evidence-based rates for sickle cell and β thalassaemia for use in the U.K., and should underpin the future planning of services. The long-term solution to monitoring changes in the rates of trait and disease in the population is to introduce a standardized instrument for collecting ethnicity for all community screening programmes.
In 1993 the U.K. Standing Medical Advisory Committee (Department of Health, 1993) recommended that pre-conceptual carrier diagnosis for the conditions should be encouraged, and that antenatal and neonatal screening should be universal in districts where >15% of the population is from ethnic minorities. However, the SMAC report, although generally welcomed (Davies, 1993), received little attention. In contrast, the Sickle Cell Disease Guideline Panel (1993) convened in the U.S.A. recommended that all states carry out universal neonatal screening for sickle cell, because of the ethical and practical difficulties raised by targeting.
Current policy on screening in the U.K. was outlined by the Chief Medical Officer in 1994 (Calman, 1994), indicating that the introduction of future programmes should depend on the evidence that particular criteria were met, as described by Wilson & Jungner (1968) and updated by the National Screening Committee (1998). These stressed that the health problem should be important, acceptable screening tests available, effective and acceptable interventions available, and the cost-effectiveness of screening demonstrated. Previous studies have shown that screening tests for haemoglobinopathies are reliable and relatively cheap (Lorey et al, 1994; U.S. Department of Health and Human Services, 1993), that for populations affected by thalassaemia antenatal screening and subsequent termination of affected pregnancies is acceptable (Modell & Bulyzhenkov, 1988), and that there are clear benefits conferred by screening and detecting infants with sickle cell in order to start life-extending prophylactic penicillin (Gaston et al, 1986). However, there has been little economic analysis of the benefits of universal versus targeted haemoglobinopathy screening and the U.K. service remains patchy (Streetly et al, 1993).
Clearly, knowledge of the number and proportion of women who carry a haemoglobinopathy and of babies with clinically significant disease is required to help haematologists and commissioners plan and evaluate the effectiveness of services and screening programmes. Previously, ethnic-specific estimates were not available for the U.K.; existing screening programmes did not collect ethnic group reliably or consistently, and the evidence supporting earlier point estimates (W.H.O., 1987, 1985) was neither clear or directly transferable to populations in the U.K.
We report here a reconstruction of ethnic-specific estimates of the haemoglobinopathies to use in the U.K., identifying the sources and assessing the strength of evidence from the literature, validating them against observed data from several local screening programmes, and finally mapping the prevalence in England.
Defining the denominator: ethnic-specific births
Birth statistics do not record ethnic group. Therefore, to estimate the distribution of ethnic specific births according to the 1991 census, proxy proportions were calculated using the number of children aged 0–4 years recorded in the census (OPCS, 1993), adjusted for under-enumeration (OPCS and GRO Scotland, 1993), and expressed as a percentage. In addition, Cypriots were identified separately from whites because of their high risk of β thalassaemia (Thalassaemia Working Party of the BCSH General Haematology Task Force, 1994).
Cypriot births were estimated by doubling the reported number of children aged 0–4 in households where the head was born in Cyprus and subtracting it from the White ethnic group. This was based on the following assumptions: only half the ‘Cypriot’ population of child-bearing age were born abroad, since most emigration occurred from 1957 to 1967; and the geographical distribution of Cypriot parents is similar whether they were born in the U.K. or Cyprus (Modell & Berdoukas, 1984). The resulting percentages were used to adjust the 1994 birth statistics by ethnic group. This was carried out for each local authority/county district in England.
Defining the numerator: ethnic specific rates of haemoglobin disorders
Estimates were derived for β thalassaemia, Hb S, C or E trait. The proportion of births with clinically significant disease (β thalassaemia major or intermedia, Hb SS, Hb SC, Eβ thalassaemia and Sβ thalassaemia) were calculated using the Hardy-Weinberg equation, where: frequency of homozygous disease = p2, with p = gene frequency of abnormal trait (≈ carrier frequency/2); frequency of compound heterozygous disease = (p + q)2− (p2 + q2), with p and q = gene frequencies of different interacting haemoglobin disorders.
Rates were adjusted for customary consanguineous marriage in Pakistanis (Darr & Modell, 1988) and for marriage outside the ethnic group for Cypriots. The weight of evidence for the estimates were graded (shown below) from being based on a population screening programme (considered analogous to a randomized clinical trial) to expert opinion (U.S. Preventive Services Task Force, 1989). Upper and lower estimates were derived if there was insufficient evidence to support a single value for an ethnic group living in the U.K.
We did not attempt to determine estimates for alpha thalassaemia because its clinically significant form is relatively uncommon in the U.K., we could not validate the estimates, and the implications for screening have been dealt with previously (Petrou et al, 1992). Haemoglobin D Punjab was also excluded, although common among Indians (Sukumaran & Master, 1974), because its most significant clinical problem (compound heterozygosity with Hb S) is rare in the U.K. and would be encompassed by the range of estimates for SCD.
Validation of estimates of prevalence of haemoglobin disorders
Estimates of the total number of affected births in each county district/local authority were obtained by combining the ethnic-specific prevalence rates and the estimated number of births by ethnic group. The estimates were validated against two London universal neonatal screening programmes (North Thames West; Lambeth, Lewisham and Southwark) and against two antenatal screening programmes in London and Leicester.
Estimates of the number of affected births in England were adjusted for termination through prenatal diagnosis (Greengross et al, 1998; Modell, 1993). A credible range was calculated from a formula for combining two or more estimates which also have upper and lower values (Day Report, 1996, Table 7, pR10). Maps of the prevalence of disease in England were produced using ARC-INFO software.
1Table I presents our estimates of the rates of carrier frequency and clinically significant disease by ethnic group for use in the U.K., including an assessment of the strength of evidence in support of them, and the main sources of information. The best data were obtained from population-screening programmes in Jamaica which were adopted for the black Caribbean ethnic group, and from Cyprus, although the latter were adjusted for the potential reduction of clinical disease in the U.K. due to marriage with non-Cypriots. Data derived from a mixture of sources for most of the other ethnic groups rarely supported a single estimate.
Table 1. Table I. Planning estimates of haemoglobinopathies by ethnic group for people living in the U.K.
* Grading: A: based on large-scale population survey in the U.K.; B: based on large-scale population surveys in country of origin with clear links to population living in the U.K.; C: Expert advice based on range of studies in country of origin with support from studies in the U.K.; D: expert advice based on unpublished data; E: assumed to be the same as another ethnic group. 1 Lower and upper estimates given where insufficient evidence to supply a single estimate for population living in the U.K. 2 Including beta thalassaemia major, and Bthal E, excludes homozygous E. 3 Including homozygous sickle, SC and SBthal, excludes homozygous C. 4 High estimates combine high AS and low AC rates; low estimates combine low AS and high AC. 5 Black other assumed to the same as black Caribbean. 6 Allowing for consanguineous marriage (of half between first cousins) increasing probable rate of homozygotes by 2. 7 Haemoglobin E assumed to be 4% for all three estimates, included in rates of compound heterozygous disease: Bthal/E. 8 Central estimates assume reduction in homozygotes due to partner exchange with non-Cypriot of 20% (lower bound = 40% and upper = none). 9 Other-other assumed to be the same as background rate in white population.10 General background references for all groups (W.H.O., 1985, 1987, 1988; Hereditary Disease Programme, 1994; Brozovic & Stephens, 1991; Model et al, 1992; Livingstone, 1985).
Sickle cell was concentrated within black ethnic minorities, with comparatively high carrier rates responsible for high rates of disease: 5.6 per 1000 births among black Caribbeans and 14.7 among black Africans. The upper and lower range for sickle cell and Hb C among black Africans reflected differences between countries in Africa. For example, Nigeria had relatively high rates of sickle cell and low rates of Hb C, whereas Hb C was more prevalent in Ghana.
In contrast, β thalassaemia was present in all populations living in the U.K., including trace amounts within the autochthonous white population, but at lower rates with the exception of Cypriots. The range of estimates for β thalassaemia for Indians, Bangladeshis, Pakistanis and Chinese reflected both uncertainty over the true population value and heterogeneity within their countries of origin, because of the lack of good applicable studies or screening programmes.
2Tables II and 3III compare the observed annual number of carriers and affected births identified by local English neonatal and antenatal screening programmes to the numbers derived from our estimates shown in Table I. Approximately 13% of births affected by sickle cell disease (SCD) were terminated in the two London programmes (North Thames West, NTW; Lambeth, Lewisham and Southwark, LLS). 2Table II shows that in NTW as a whole and in almost all of the individual health authorities the observed number was within the expected range, and close to the central estimate. In LLS the observed number of SCD and haemoglobinopathy trait was within the expected range but towards the upper estimate (Table III). In both Leicester and Brent the observed number of β thalassaemia carriers were closer to the central estimate than the upper or lower range (Table III).
Table 2. Table II. Observed and expected number of births and infants with sickle trait and disease: North-West Thames universal neonatal screening programme.
1 Hammersmith & Fulham, Kensington & Chelsea, Westminster. 2 Dr Old (personal communication).
Table 3. Table III. Observed and expected number of births and pregnancies: sickle cell disease and haemoglobinopathy trait in four areas.
1 Includes six (13%) cases diagnosed through prenatal diagnosis and terminated.2 Served by King's College; estimated number of trait adjusted to number of deliveries at King's College.3 Target Asian women and their partners (approx. 30% of total births and bookings).4 Estimated number of affected births adjusted to number of births booked at Central Middlesex Hospital by borough of residence.
Finally, there were 22 cases of β thalassaemia major/intermedia (64% terminated) identified in NTW from 1990 to 1994. Over a similar time period, using our ethnic-specific estimates of births and haemoglobinopathies, we estimate that there would be 19 cases (lower to upper range 12–27), which also was a close approximation.
Our validated estimates (Table IV) suggest that each year in England about 3000 (0.47%) babies are born with sickle cell trait and 2800 (0.44%) with β thalassaemia trait, i.e. nearly 1 in 200 each. Annually approximately 178 (0.28 per 1000) conceptions will be affected by SCD and 43 (0.07 per 1000) by β thalassaemia major or intermedia. Allowing for selective termination of 50–70% of births with β thalassaemia and 5–15% of SCD, we estimate that 140–175 (0.22–0.28 per 1000) affected infants are born annually with SCD and 10–25 (0.02–0.04 per 1000) with β thalassaemia major/intermedia.
Table 4. Table IV. Estimate of number of pregnancies and births affected by β thalassaemia or sickle cell in England.
1 Including beta thalassaemia, and β-thal E. 2 Including homozygous sickle, SC and Sβthal.3 Range: total estimate − √Σ(central estimate − lower)2 to total estimate + √Σ(central estimate − upper)2, where the sum is over the central, lower and upper estimates of disease prevalence and % termination.
The distribution of ethnic minorities was geographically highly skewed, with the proportion living in individual districts ranging from 0% in the Isles of Scilly to > 60% in Brent, with half of the districts having < 3%. Figs 1 and 2 show the geographical distribution of disease for SCD and β thalassaemia, respectively, by county district in England. These highlight the heterogeneity in prevalence and the clustering in inner-city areas with higher proportions of ethnic minority populations, and the importance of concentrating on cases of disease rather than carrier frequency. β thalassaemia trait was as prevalent as sickle cells and geographically was more widespread because of its presence in the White population. However, 19 county districts (5% of the total) would be likely to experience a case of β thalassaemia major or intermedia within 2 years; far fewer than the 51 (14% of the total) that would be likely to encounter a case of SCD.
Our estimates simplify the composition and pattern of haemoglobin disorders in the U.K. population. First, some population groups without routine age-specific data and information on migration were not separately identified, though at greater risk than the autochthonous white population, such as sickle cell among southern Italians and other Mediterranean countries (Silvestroni et al, 1968; Cao et al, 1989). Second, some rare haemoglobin disorders have not been included. Third, apart from Cypriots, no allowance was made for ‘marriage’ between ethnic groups in estimating rates of disease, which may increase or decrease the risk of having an affected fetus. Increasing inter-marriage will invalidate these estimates in the future as well as reduce the sensitivity of targeted screening (Adjaye et al, 1989; Barton & Watson, 1988). Fourth, the estimates of disease assume that people having two pregnancies within one calendar year have the same probability of having an affected child as those who only have one baby. This may not be the case for women who undergo a termination after a positive prenatal diagnosis, and may slightly under-estimate the number of babies affected by thalassaemia, although this was not detected in our validation exercise of North West Thames.
Nonetheless our estimates (shown in Table I) are the best available data on the prevalence of haemoglobin disorders for use in the U.K. and offer both an evidence base and a methodology for other countries. When combined with estimates of the number of births by ethnic group (which can be derived as suggested by our methods) they provide important data on the likely number of affected births to enable haematologists and health authorities to plan screening and treatment services. Moreover, they are the first evidence-based rates for sickle cell and β thalassaemia for use in the U.K. They take forward and replace earlier unverified point estimates for the U.K. such as those used in the report on ethnicity and health by NHS Centre for Reviews and Dissemination (Modell & Anionwu, 1996), and have been adopted by more recent studies (e.g. HeA, 1998, and Zeuner et al, 1999).
There was a lack of evidence to support single values for many of the ethnic groups. Therefore a range of values were provided reflecting the heterogeneity of prevalence within specific ethnic groups and which were fully validated by being able to predict the observed prevalence in local districts. For example, both Brent in NTW and Lambeth, Lewisham and Southwark (LLS) have a high proportion of ethnic minorities, but in Brent the central estimate was close to the observed data whereas in LLS the upper estimate was the nearest to the observed data, probably because more black Africans living in LLS originated from countries with high carrier frequencies for HbS, such as Nigeria (M. Layton, personal communication).
Ethnicity is clearly a good marker of the populations at risk from haemoglobinopathies, as long as Cypriots are included. However, previous expert advice given to haematologists and health-care planners to commission universal or targeted screening on the basis of the proportion of ethnic minorities in the population is over-simplistic and misleading (Department of Health, 1993), and cannot be resolved simply by lowering the cut-off for universal screening from 15% to 10% (Modell & Anionwu, 1996).
Firstly, it is inequitable. A different composition of black and other ethnic minorities can produce different numbers of affected births, and inevitably a single cut-off value (based on percentage of ethnic group) could exclude areas with similar or higher rates of haemoglobinopathies than areas above the cut-off. Secondly, it rests on the false assumption that the proportion of ethnic minorities is equivalent to cost-effectiveness. Other work, at least for neonatal screening, suggests that the volume of tests are crucially important in determining cost-effectiveness, and that programmes need to screen >25 000 babies annually (Cronin et al, 1998) which is 5–10 times greater than the number of babies delivered in most district general hospitals in England and clearly has other implications for the way screening services are organized.
In conclusion, the best way of improving the evidence base and obtaining better data (both on the numerator, carrier rates, and denominator, conceptions and births) will be through the introduction of a standardized instrument for collecting ethnicity data. Thus, data could be combined from the existing and future community neonatal and antenatal haemoglobinopathy screening programmes to establish and monitor changes in the rates of carrier frequency and disease in the population. We recommend early development of this work.
We thank Dr Mike Gill, Director of Public Health and Health Policy Brent and Harrow HA, for his support and encouragement. We also thank Dr Old for supplying the data on prenatal diagnoses in North Thames RO. This work was supported by the NHS Technology Assessment Programme grant 93/33/3.