An overview of concepts and approaches used in estimating the burden of congenital disorders globally

Congenital disorders are an important cause of pregnancy loss, premature death and life-long disability. A range of interventions can greatly reduce their burden, but the absence of local epidemiological data on their prevalence and the impact of interventions impede policy and service development in many countries. In an attempt to overcome these deficiencies, we have developed a tool—The Modell Global Database of Congenital Disorders (MGDb) that combines general biological principles and available observational data with demographic data, to generate estimates of the birth prevalence and effects of interventions on mortality and disability due to congenital disorders. MGDb aims to support policy development by generating country, regional and global epidemiological estimates. Here we provide an overview of the concepts and methodological approach used to develop MGDb. Electronic supplementary material The online version of this article (10.1007/s12687-017-0335-3) contains supplementary material, which is available to authorized users.


Introduction
Congenital disorders, often also called birth defects, include Bany potential pathological conditions arising before birth, whether they are evident at birth or become manifest later in life^ (World Health Organization 1985a;World Health Organization 2000;World Health Organization 2010). Risk factors for congenital disorders include genetic, environmental and wider societal factors. The occurrence and severity of specific congenital disorders are differentially influenced by these risk factors, with some disorders influenced more by genes (e.g. single gene disorders) and others by environmental agents (e.g. those caused by infections). They are an important cause of premature death or life-long disability; however, the This article is part of the Topical Collection on Epidemiological Methods in Community Genetics Electronic supplementary material The online version of this article (https://doi.org/10.1007/s12687-017-0335-3) contains supplementary material, which is available to authorized users. absence of local epidemiological data on their birth prevalence and outcomes impedes policy and service development in many countries.
To make up for this deficiency, the Modell Global Database of Congenital Disorders (MGDb) was created to meet the information needs of health policy-makers. It does so through the generation of country, regional and global estimates for selected congenital disorders, by combining general biological principles and available observational data with demographic data. Here we describe the scope of MGDb and the general principles followed in its implementation; more detailed methodology can be found in other articles within this journal supplement and in the UCL repository .

Terminology
MGDb uses a set of defined methods to relate demographic data to the known birth prevalence of selected congenital disorders, in order to generate estimates relevant to public health, policy-making and clinical practice. The creation of this tool requires precise definitions of all input data and outputs. Table 1 provides a list of some of the key terms along with their definitions within the context of this tool; a comprehensive glossary can be found in the UCL repository ).

Scope of conditions
Only severe, early-onset congenital disorders that cause early death and/or life-long disability in the absence of care and present before 20 years of age are included in MGDb. The disorders currently modelled by MGDb are shown in Table 2. The rationale for modelling these disorders, where data are unavailable, is that their birth prevalence in the absence of access to diagnosis and care (baseline birth prevalence) is relatively constant in any given population. Estimation of the birth prevalence of these conditions can therefore be undertaken assuming that this will vary only with identified factors, such as interventions (e.g. folic acid fortification, prenatal diagnosis, surgical repair). Modelled estimates can be generated on a country-specific annual basis of potential (in the absence of interventions) and actual affected births. Thus, the tool can also be used to estimate the actual effects of any proposed future interventions. Although the same method can Effective cure A disorder that has been sufficiently corrected to allow affected individuals to live their lives free from continuing medical care, and to achieve life goals such as independent living, finding a partner, reproductive success, even with some persisting problems. It does not mean complete correction with no residual problems be applied to congenital disorders that arise as a result of external risk factors, such as maternal exposure to infection, malnutrition, or teratogens, at present there are insufficient input data at the global level to define a baseline birth prevalence and outcomes of these disorders. This is due to lack of precise estimates of risk associated with these exposures and the fact that risk varies more widely with place and time, requiring country-specific data.
The congenital malformation group includes the International Classification of Disease (ICD 10) (World Health Organization 2016) system groups used in most congenital anomaly registries and focuses on isolated malformations (i.e. not those associated with chromosomal disorders or genetic syndromes. Disorders associated with other malformations contribute to the multiple malformations group. Neural tube defects, orofacial clefts and congenital heart defects, are treated  Bundey and Alam (1993), Bittles and Neel (1994).  Kromberg et al. (1989), Lund and Taylor (2008) Genetic risk factors Rhesus haemolytic disease Mourant et al. (1976), Bhutani et al. (2013) G6PDd kernicterus World Health Organization (1985b), Howes et al. (2012) a CHD congenital heart defect that usually presents before 20 years of age and would cause premature death or disability in the absence of intervention b Hypothyroidism due to thyroid agenesis or dysgenesis. Hypothyroidism due to iodine deficiency is excluded c Recessive disorders that would occur in the absence of consanguineous marriage and are disadvantageous without any identified compensating selective advantage d Parental consanguinity is associated with an increment in congenital disorders, the increment is mainly due to increased birth prevalence of recessive single gene disorders as parental consanguinity increases the chances that a couple will both carry the same recessive gene variant separately, as data relating to these malformations are more readily available. The chromosomal disorder group includes all chromosomal disorders that cause substantial disability for the affected person. In MGDb, Edwards and Patau syndrome are treated together as Bother trisomies^because their outcomes are very similar. Inherited disorders have been bundled into four broad groups. Rare single gene disorders include dominant, X-linked and recessive disorders that are expected to have a similar collective birth prevalence worldwide, because their gene frequency is mainly determined by the balance between new mutation rate and natural selection, neither of which is thought to vary greatly between populations. The common recessive disorders group includes three conditions (sickle cell disorder, thalassaemia and oculocutaneous albinism) for which there is known global variation in carrier prevalence. Consanguinity-associated disorders refer to the increment of recessive disorders that are associated with parental consanguinity (Corry 2014). Disorders due to genetic risk factors are caused by the interaction of (often very common) DNA variants with other genetic and environmental factors. MGDb includes two early-onset examples (rhesus haemolytic disease of the newborn and neonatal jaundice due to G6PD deficiency) because (a) the underlying mechanisms are exceptionally well understood and (b) they are potentially lethal but can be effectively prevented and/or treated.

Overview of the methodological approach
The objective of MGDb is to estimate numbers of births affected by one or more congenital disorders, and outcomes in the no-care situation and with current care. Figure 1 shows the range of possible outcomes and their modification by a variety of diagnostic and therapeutic interventions. The effect of some interventions is quantifiable as data relating to their impact are available, for example, folic acid food fortification, the identification of genetic risk and the option of termination of pregnancy. In addition to the impact of particular interventions, their availability and ease of access must be taken into account (Blencowe et al. 2017). Estimates were derived in a stepwise manner, beginning with baseline birth prevalence and outcomes in the absence of interventions (Fig. 1). Once an estimate is available for baseline birth prevalence, outcomes may be calculated based on estimated actual live birth prevalence, estimated access to services and survival with optimal or very limited care.

Sources of data Demographic data
Demographic data not only provide the basis for quantitative epidemiological estimates and the assessment of service needs, they offer an overall picture of many aspects of a population. It is particularly important to recognise the speed of demographic change; all epidemiological estimates need to take account of this dynamic background. Health Metrics and Evaluation (IHME) 2015; Inter-agency Group for Child Mortality Estimation (UN-IGME) (2015); UNAIDS 2016). Table 3 provides a summary of the demographic indicators and their sources used to generate estimates for the prevalence and outcomes of congenital disorders.

Baseline birth prevalence
The frequency of a disorder in a population is usually described in terms of incidence (number of new cases arising in a defined time-period) and prevalence (number of cases actually present in a given unit of population, e.g. per thousand or per million). However, the incidence of congenital disorders is usually expressed as birth prevalence because although many congenital disorders arise at or around the time of conception, many affected embryos either fail to implant or miscarry in early pregnancy, and so never come to the attention of health services. Therefore, for practical purposes, their prevalence at birth is counted as their incidence, and expressed as rate per 1000 live births. MGDb follows the European Surveillance of Congenital Anomalies and Twins (EUROCAT) convention in counting all pregnancy outcomes after 20 weeks gestation as births. In keeping with ICD-10, all births with any signs of life following separation from the mother, regardless of gestation, are counted as live births. Baseline, or potential, birth prevalence is the prevalence that would apply in the absence of any intervention. It is the foundation of the MGDb and provides the Benvelope^into which all outcomes must fit. It includes stillbirths and live births but excludes miscarriages, and uses Bfoetal death( death in utero after 20 weeks' gestation) as a proxy for stillbirth. Once baseline birth prevalence is known, when appropriate observational data are available baseline births can be allocated to each potential outcome ( Fig. 1) thus generating an overall epidemiological picture of the current status of each disorder group.
The baseline birth prevalence for most countries worldwide can be calculated from available data or estimated for many congenital disorders ( Table 2). The methodology used to obtain these estimates is described in detail elsewhere (Modell  Piel et al. 2013a, b). With regard to congenital malformations, there are sufficient observational data available in the literature and from registries on neural tube defects and orofacial clefts to make estimates for countries in every region Blencowe et al. 2017), since these severe malformations are evident at birth. However, data on all other congenital malformations and rare single gene disorders are only available in high-income settings. To generate estimates for congenital malformations for countries outside high-income settings, we have utilised publicly accessible data from the EUROCAT website to obtain average European rates for birth prevalence (Moorthie et al. 2017;Modell et al. 2016). Studies examining ethnic differences have only shown variation for polydactyly, congenital hypothyroidism, orofacial clefts and neural tube defects (Chitty and Winter 1989;Petrini et al. 1997;Terry et al. 1985;Yang et al. 2004). In addition, data from the International Clearinghouse for Birth Defects Surveillance and Research (ICBDSR) indicate little inter-ethnic variation for other malformations. As a result we have assumed that the baseline birth prevalence for these conditions is little affected by ethnicity; therefore, average European rates have been used to generate estimates for countries with no or insufficient observational data ). In the absence of more recent data on the birth prevalence of rare single gene disorders, we have utilised the rates of Baird et al.  to represent global collective baseline birth prevalences. However, the contribution of specific single gene disorders to collective birth prevalence is known to vary in countries, especially where there are founder effects.
Data on outcomes in the absence of diagnosis and care, and with optimal care (i.e. with the interventions generally available in high-income settings), can be obtained for most groups of congenital disorders. In any setting, a proportion of the population has access to optimal care; however, there is no routinely collected indicator from all countries that provides this information. We have therefore developed a method for estimating the proportion of access to optimal care using infant mortality as a proxy indicator . Country-specific estimates of outcomes can then be calculated, based on outcomes in the absence of care and with optimal care, and the proportion of the population with access to care ).
There are only three possible outcomes for early-onset congenital disorders in the absence of interventions-foetal death, life with some degree of disability and premature death (Fig. 1). The types and availability of specific interventions have evolved over time, impacting on birth prevalence for specific disorders in different time-periods and the subsequent number of survivors . Increased access to prenatal diagnosis and termination of pregnancy can act to reduce birth prevalence whilst improvements in diagnosis and care impact on outcomes. Provision of early diagnosis and care can greatly reduce annual numbers surviving with disability for disorders with effective cures. For children with incurable disorders it can both prolong survival and ameliorate the levels of disability. As a result, there may be a steady annual increase in the cumulative number of individuals living with these disorders and requiring appropriate care. The evolution of this effect needs to be quantified in order to assess current and future patient numbers and service needs. A full description of the sources of data on the impact of interventions is available elsewhere (Blencowe et al. 2017;Modell et al. 2016).

Under-5 mortality
For most congenital disorders mortality is highest within the first 5 years of life, the period for which survival data with no care and with optimal care are most complete and most reliable. To avoid double counting, it is important to allow for the overlaps inherent in multiple causes of death. For assessing total deaths of people with a given disorder, all deaths of affected individuals must be included, whatever the cause. However, when the aim is to assess attributable deaths-deaths that are specifically due to a defined disorder, the background mortality should also be considered. This is because some affected individuals who would have died of their disorder in fact die earlier from unrelated causes, and the proportion of such deaths varies with place and time.
To obtain attributable early deaths, numbers are adjusted for deaths from other causes using country rates for neonatal, infant and under-5 deaths ). The adjustment makes relatively little difference in countries where early background mortality is low, but when background mortality is high a sizeable difference is observed. Table 4 shows the main sources used for estimating survival with limited care, and with optimal care, defined as the best care available around the time the patients were born. The upper age limit of the observational data is also shown. This allows the construction of survival curves in a no-care and an optimal care situation which were then used to estimate number of survivors.

Long-term survival
Lifetime survival curves with optimal care are only available for some congenital disorders-Down syndrome, spina bifida, orofacial clefts, and haemoglobin disorders. For most other disorders, observational survival data are only available to age 20 or 30 years. Lifetime survival curves were completed for these disorder groups by extrapolating the observed rate of attrition in the last full 5-year interval recorded to 80 years of age. Normal survival is expected for correctable congenital malformations, congenital hypothyroidism and disorders due to genetic risk factors, as optimal care should allow for effective cure or appropriate management of these conditions. Long-term survival in the absence of care is based on literature available from high-income settings in the 1950s for spina bifida, Down syndrome, congenital heart disease and haemoglobin disorders and on expert opinion for most other disorders (Table 4).
Lifetime survival curves are used to project future patient numbers, but they need adjustment for use in calculating  Macmahon and McKeown (1953), Campbell (1968), Campbell et al. (1957), Baylis and Campbell (1956) Severe CHD current patient numbers, since mortality is usually higher in the early years. The current number of survivors in each age group is largely determined by the services available around their year of birth. Retrospective survival curves were derived from prospective curves, taking into consideration the historical evolution of services, for use in calculating the number and age distribution of patients living at any given time. For this purpose, survival data for each phase in the evolution of care were obtained from the literature .
Disability and cure with optimal care and with no care Definitive cure of conditions is limited to operable congenital malformations. Survivors with most other disorders live with some disability. The survival outcome can range from well (e.g. congenital hypothyroidism with regular replacement treatment) to the very severe (e.g. associated severe physical and mental disability). Due to this diversity, quantification of physical disability is limited to estimates of the proportion of survivors at age 5 with severe disability (including reduced life expectancy), less severe disability (with lesser or no effect on life expectancy) and effectively cured. Data regarding long-term disability outcomes, depending on the levels of access to care, were obtained through literature review. The proportion of each population without optimal care is assumed to have no access to care. Although we have used the term no-care, this does not refer to the absolute lack of care as this is unlikely to be the case in reality, with supportive care at home or in a facility available in many settings. In addition, we have used estimates for survival with no care based on available data, which in large part is based on survival in the 1950s in Bhigh income^settings ); consequently the survival that we have estimated is likely to be an overestimate for a total absence of care.

Examples of MGDb outputs by WHO region
The aim of the present exercise is to support health policy makers and practising clinicians at the country level.  Fig. 3 Estimated outcomes of congenital disorders if no intervention, by type of outcome, 2010-2014. Note: The small discrepancy between total affected birth prevalence in the previous chart and total outcomes represents estimated under-5 deaths due to other causes (omitted for the sake of clarity). AFR African, AMR American, EMR Eastern Mediterranean, SEAR South-East Asian, WPR Western Pacific Region, W Europe Western Europe Outputs are therefore generated to provide a basis for assessing non-financial and financial costs and benefits, and to allow critical comparison with other estimates. MGDb output data are generated for each disorder group by country; these can be aggregated by WHO region. All outputs are expressed in terms of annual numbers and rate per 1000 live births.
Baseline birth prevalence and actual birth prevalence Figure 2 shows the estimated baseline birth prevalence of the groups of congenital disorders by WHO region. All charts also include rates for Western Europe because most currently available interventions are deployed in this sub-region near-equitably at high coverage, and surveillance data are available. Observed outcome data may therefore be used to describe the Bpower^of each intervention when fully deployed at the population level. There is little inter-regional difference in the baseline prevalence of chromosomal disorders, congenital malformations and baseline single-gene disorders. Most of the inter-region difference is due to genetic disorders where there are inter-country differences in carrier prevalence, or according to the prevalence of consanguinity. Figure 3 shows the corresponding distribution of three possible outcomes in a situation with limited access to care, providing a baseline for assessing the effects of interventions. The estimated effect of pre-birth interventions on affected birth prevalence in 2010-2014 by type of intervention can be seen in Fig. 4. Globally, the largest contributions were from anti-D for rhesus negative mothers and termination of pregnancy for foetal impairment. The effect of pre-pregnancy genetic counselling is relatively small because in most cases risk is only detected retrospectively, i.e. after the diagnosis of an affected child .

Birth outcomes, outcome at 5 years of age and effect of interventions
The estimated distribution of actual outcomes in 2010-2014, taking account of interventions and charted within the envelope of potential birth prevalence, can be seen in Fig. 5. By definition, the sum of all outcomes fits into the envelope of baseline birth prevalence. Rates for Western Europe indicate that all available interventions, when deployed at the population level, reduce mortality due to congenital malformations, chromosomal disorders, the two genetic risk factors and  haemoglobin disorders by over 80%, but there is far less effect on mortality due to other single-gene disorders. Effective cure is possible for many congenital malformations by paediatric surgery (online resources Fig. 1). Table 5 shows the estimated per cent reduction in unfavourable outcomes due to interventions at 5 years of age, in 2010-2014. The greatest proportional reduction is in under-5 deaths. There is much less reduction in disability-in fact, disability at age 5 due to congenital disorders was increasing in the Eastern Mediterranean, African and South-East Asian regions.
Mean life expectancy Table 6 shows disorder-specific mean life expectancy with no care and optimal care, calculated from the full survival curves.
The difference between life expectancy with no care and with optimal care measures the survival benefit of care. Mean life expectancy is used to calculate costs of the disorder and the benefits of interventions, in terms of years of life lost, lived with disability or lived cured. The rates in Table 6 represent life expectancy with the disorder in the absence of any other cause of death. These Bideal^rates are adjusted by multiplying local life expectancy divided by a notional optimal life expectancy of 80 years. Years affected per person born is the same as local mean life expectancy. This provides the envelope for all estimates of outcomes. The world average is 2.5 years affected per person born (range 3.25 for the EMR to 2.2 for WPR). In the absence of intervention, they would be responsible for 2 years loss of life per person born (range 1.7-2.7 years). Figure 6 shows estimated outcomes in terms of years of life lost, lived with disability or lived cured per person born in 2010-2014. The striking reduction in death and disability in Western Europe demonstrates the potential of global implementation of interventions for congenital disorders.

Estimated number and age distribution of living patients
To plan for current service needs and project future changes, policy-makers need to know the approximate numbers of persons presently living with specific disorders. This information can only be reliably obtained using a patient register. However, few such registers exist even in high-income  settings, even for readily diagnosable disorders such as Down syndrome. The method we have developed extends the principles described above (i.e. consideration of availability of interventions, their impact and evolution along with demographic data) to estimate the number and age distribution of patients currently receiving care. As WPP provides age distribution data from 1950 onwards, it is possible to generate estimates in any year from 1950 onwards, allowing a comprehensive picture of the history of selected disorders up to a chosen year. As an example, Fig. 7 shows global estimates for spina bifida by 5-year age intervals in 2010.

Future projections
Future projections of the likely effects of implementing interventions, such as folic acid food fortification or prenatal diagnosis with the option of termination of pregnancy are of particular interest to policy-makers. They can be calculated as follows: (a) assuming no change in present policies and (b) assuming worldwide spread of available interventions. Table 7 summarises the estimated long-term effects of different policy decisions for spina bifida at a global level. All of the estimates take account of change in access to care during the time period involved.

Discussion
The absence of epidemiological data on congenital disorders has led to the creation of MGDb. However, as with all modelling initiatives, the estimates produced can only be as accurate as the input data, and they are approximations only. We emphasise that our aim is only to produce order of magnitude estimates that can be used as a starting point for service planning. They must not be considered as definitive, and require continuous improvement and refinement as better information becomes available. In undertaking this task, it proved necessary to clarify some concepts and develop some new methods  to address gaps in available observational data. Articles within this journal describe some methodological aspects and further details can be accessed via the UCL repository .
Most estimates generated by MGDb are likely to be underestimates. This is because it is difficult to diagnose many congenital disorders, especially when they cause stillbirth or early death, or in settings with limited diagnostic capacity, which can lead to death and disability being attributed to other causes and a lack of reliable prevalence data. Reliable cause of death data can only be obtained in high-income settings where, by definition, prevention and care are also available. Therefore, if medically certified death rates are assumed to apply globally, early deaths due to congenital disorders will be greatly underestimated. In addition, wherever possible attempts were made to avoid overestimation; some examples are given below. The estimates for single gene disorders are based on older literature and hence only cover conditions diagnosable at that time; but with advances in DNA technology new single gene disorders are being steadily identified (Boycott et al. 2013). The rates for effect of parental consanguinity are largely based on the work of Alan Bittles; but this largely estimates mortality rather than mortality plus disability (Bittles 2003;Bittles and Black 2010), and rates for Rhesus haemolytic disease are lower than those published elsewhere (Bhutani et al. 2013).

Strengths
The strength of this approach lies in beginning with baseline birth prevalence, which provides the most robust estimates. It provides the envelope for all possible outcomes; consequently, overestimation of one outcome leads to underestimation of another, for example overestimation of mortality leads to underestimation of disability. We have endeavoured to address factors that can lead to overestimation of mortality and prevalence as detailed by Liu et al.(Liu et al. 2012 and described below. 1. Potential bias towards high prevalence populations, which is a possibility when rates are based on the literature. However, we have endeavoured to use rates based on international registry data. For high-income settings the rates come from EUROCAT, which contains populationbased data. Most data for lower-income settings has been obtained from ICBDSR, and most participating registries are hospital-based. This could result in bias due to selective referral of high risk pregnancies, but is unlikely as this often requires routine foetal anomaly scanning. Gene frequency data for haemoglobinopathies, G6PD deficiency and rhesus negativity are based on many large-scale population surveys. Where data were from selected populations, they were excluded. 2. The uncertainty in applying prevalence estimates derived from largely European populations to populations with different fertility patterns, or specific genetic and environmental contexts. For congenital malformations, comparative studies by ethnic group in high-income settings yield comparable rates with two exceptions: (a) a lower birth prevalence of neural tube defects and orofacial clefts among people of Sub-Saharan African origin, and (b) an increased prevalence of congenital malformation syndromes in groups in which consanguineous marriage is customary. Allowances have been made for differences in maternal age distribution, total fertility rate and observed ethnic differences in the birth prevalence of NTD, OFC, CHT and the effect of parental consanguinity by using country-specific data for these conditions and estimating the impact of consanguinity. FAF folic acid fortification, TOP termination of pregnancy 3. The fact that some estimates include stillbirths and terminations as well as live births. Due to differences in the definition of stillbirths we have used foetal deaths as an approximation for stillbirths. As many terminations occur before 20 weeks, we have adjusted TOP rates to those that would apply from 20 weeks onwards. In addition, when observed data applied only to live births we added estimates for foetal deaths based on data from registries and the literature. 4. Double counting of infants with more than one congenital disorder. We ensured throughout that numbers and rates apply for affected individuals within the groups specified in Table 1 and created a separate category for cases with more than one congenital malformation. 5. Overestimating mortality due to congenital disorders in lower income settings. Estimates for mortality in a nocare situation are based on historical data from highincome settings and are likely to underestimate a truly no-care situation.

Weaknesses
As improved data become available from other regions, it will be important to test the assumptions made and to modify them as necessary. The biggest assumptions relate to our estimate of access to care and long-term survival, due to the paucity of data in these areas. A further limitation of this work is a paucity of population-based, cause-specific fatality rate data for varying care settings, necessitating reliance on historical data or expert opinion. There is an urgent need for improved outcome data to verify or amend these estimates. This is an important area for future research, to fully document the longterm impact of living with these conditions and improving the estimates available to more fully quantify the benefits arising from interventions.

Differences from other estimates
Currently, the only available national and regional estimates for mortality and disability due to congenital disorders were undertaken as part of the Global Burden of Diseases (GBD) study. GBD estimates are limited to congenital anomalies, defined as disorders included in ICD10 chapter VII (the Q chapter)-BCongenital malformations, deformations and chromosomal abnormalities^; that is, they deal only with developmental structural anomalies. Congenital anomaly groups specified in the GBD are Down syndrome, unbalanced chromosomal rearrangements, neural tube defects, orofacial clefts, congenital heart anomalies and other congenital anomalies. These exclude single-gene disorders or genetic risk factors modelled in MGDb; consequently, they were excluded from the comparison. Table 8 summarises the results of a comparison of estimates of under-5 deaths produced by the GBD (Lopez et al. 2006b) and MGDb. There is reasonable correspondence between estimates for high-income regions and for East and Central Europe where vital registration data are available, but with wide divergences in other regions. The MGDb estimates are consistently higher for most other regions; however, they do fall within the uncertainty range of the GBD data for North Africa and the Middle East and most of sub-Saharan Africa (see online resources Fig. 2).
In existing global estimates, confusion about terminology leads to under estimation of the true burden of congenital disorders. For example, estimates of the Bcongenital^contribution to under-5 mortality published by the GBD study (Lopez et al. 2006a, b) and WHO (Liu et al. 2012) cover only congenital anomalies. They do not include genetically determined disorders (single-gene disorders and disorders due to genetic risk factors), although estimates from MGDb indicate that they contribute to around 50% of under-5 deaths due to congenital disorders. This important point underlines the need for a clearly agreed terminology for community genetics. It is widely recognised that the average baseline birth prevalence of congenital anomalies is at least 20/1000 while the baseline birth prevalence of congenital disorders (including single-gene disorders and early-onset disorders due to genetic risk factors) is over 37/1000 Czeizel and Sankaranarayanan 1984;World Health Organization 1985a). Their contribution to early death and disability is masked in lower-income settings by high early mortality from other causes, but it has been estimated that infant mortality can fall below 10/1000 only when interventions for the prevention and care of congenital disorders are in place (Christianson and Modell 2004;World Health Organization 1985a). Consequently, they would be expected to hold a significant place in the Sustainable Development Goals (to end preventable deaths in children) and the WHO strategy for noncommunicable diseases (Darmstadt et al. 2016).

Conclusions
It is possible to overcome the current difficulties in collecting high-quality population-based epidemiological data for congenital disorders in many low-and middle-income countries. The method used here to generate estimates for countries with little or no observational data makes it possible to generate useful order of magnitude estimates by (a) starting from evidence-based estimates of baseline affected birth prevalence; (b) basing estimates for high-income settings on observational data; (c) generating estimates for a baseline no-care situation using the limited observational data supplemented by expert opinion; and (d) using an empirical equation based on infant mortality rates for estimating the proportion of each population with access to the relevant services.