Examination of the frequency of congenital cardiac disease, either as a rate or as a proportion, has important implications for the study of congenital cardiac malformations, as well as their clinical management. There is much confusion and misuse, however, regarding terminology and methodology, with important implications for the accuracy, validity and comparability of findings reported in the published literature. Knowledge of how critically to appraise these reports is important in defining their value when applied to issues of diagnostic likelihood, surveillance and trends, aetiologic associations, burden of disease, and requirements for resources. These issues have more recently been impacted by fetal diagnosis and termination, with individual decisions influenced by contemporary estimates of prognosis related to the natural and modified natural history. As more patients survive into adulthood, estimates of the burden of disease, and the requirements for resources, have also achieved greater importance. The question ‘how much congenital cardiac disease?’ therefore continues to evolve. Providing the correct answer has important relevance for both the providers of health care, and the health care system itself.
DEFINITIONS
When considering the frequency of congenital cardiac disease, strict and explicit definitions are important, but often misunderstood and misused. Critical terms to be defined include congenital cardiac disease itself, frequency, ratio, proportion, rate, incidence, and prevalence.
In looking at reports of congenitally malformed hearts, it is important to know how the lesions were defined, and what conditions were included or excluded. Congenital cardiac disease has been defined as the presence of ‘a gross structural abnormality of the heart or intrathoracic great vessels that is actually or potentially of functional significance’. 1 The additional implication of this definition is that these abnormalities arose at the time of cardiovascular development, and should therefore be present at the time of delivery or, with the advent of fetal assessment, fetal diagnosis. This definition would exclude normal variants that would be of no functional consequence, such as anomalies of the systemic veins, for example, persistent patency of the left superior caval vein, or abnormal patterns of branching of the systemic arteries that have no functional consequence. Using this definition, however, does not end all controversy. There is no current consensus as to whether several groups of lesions should be considered to represent congenital cardiac malformations:
- •
Genetic conditions, obviously present from conception, which may not have manifest cardiovascular consequences until much later in life. Examples would include Marfan syndrome, Williams syndrome, hypertrophic cardiomyopathy, and pulmonary hypertension.
- •
Arrhythmic conditions with abnormalities at the physiologic or ultrastructural level, such as long QT syndrome and abnormal pathways producing ventricular pre-excitation pathways.
- •
Primary cardiomyopathies with a genetic or metabolic aetiology. This might also include myocardial abnormalities such as ventricular non-compaction.
- •
Structural defects which do not have functional significance in many, but not all, circumstances, such as the aortic valve with two leaflets, the prolapsing mitral valve, so-called silent persistent patency of the arterial duct or small septal defects, including patency of the oval foramen. This group might also include structural defects which resolve without ever becoming clinically manifest, such as small muscular ventricular septal defects. Consideration of these lesions is important because they are common, and might inflate a prevalence estimate.
- •
Persistent patency of the arterial duct in premature neonates.
Comparability of previous reports of congenital cardiac malformations has suffered from a lack of common and accepted nomenclature regarding their morphological and functional description, developmental aspects, and interrelationships. To some extent, there has been important work in consolidating nomenclature towards an internationally accepted standard. 2–4 This will greatly facilitate future collection of data relevant to incidence, prevalence, and prognosis. It does not resolve the current dilemma.
RATIOS, RATES, AND PROPORTIONS
Confusion and misuse regarding the terms incidence and prevalence is rooted in misunderstanding regarding what constitutes a ratio, a rate, and a proportion. A ratio has a numerator and a denominator, which are mutually exclusive. An example would be the ratio of males to females born with transposed arterial trunks. It has use when one is interested in the relative frequencies of two related categories, and is often represented as an entity called odds. A rate has both a numerator, indicating frequency of occurrence of a condition or event, and a denominator, indicating the time period over which the numerator was accumulated. Often the rate would apply to a defined population at risk for the condition, or an event specified in the numerator. An example would be the rate of the occurrence of a thrombotic episode in patients during their first year subsequent to construction of the Fontan circulation. Such a rate may give the impression that the condition or event occurs uniformly over the specified period of time, which may or may not be true. Rates are often used when assessing the risk of occurrence of a condition or event, and are often accompanied by explorations of factors that might be associated with increases or decreases in that rate. An often more appropriate method for looking at rates would be the calculation of Kaplan-Meier estimates, or parametric modeling of the time-related hazard, which would depict and reflect instantaneous changes in rates and risk over time. A proportion , likewise, has a numerator, which represents the frequency of a condition or category within the population specified in the denominator. It is frequently expressed as a fraction or percent, and is independent of any period of time. An example would be the proportion of patients with deficient ventricular septation in whom the defect itself is encased within the muscular ventricular septum. It would most frequently be used when describing the frequency of a characteristic or condition in a specified population.
INCIDENCE
This entity is a rate, and is strictly defined as the number of new occurrences of a condition or event identified over a specified period of time related to the number of the population at risk. The precondition is that the population at risk starts the period of observation with no occurrences of the condition or event. To apply this definition to individuals with congenitally malformed hearts, it would be necessary to begin the evaluation at conception, and then to follow embryos and fetuses throughout cardiac development to determine the number who develop defects. This would be an impossible task, since many conceptions may terminate spontaneously before cardiac development is complete, or even before conception has become manifest, or before a defect could be detected with current technology. This inability creates important challenges in the study of aetiology. Incidence and rates, however, are relevant in the study of prognosis and the natural and modified natural history.
PREVALENCE
This feature is a proportion, and is strictly defined as the number of pre-existing and new occurrences of a condition or event identified in a population at risk, either at a single point in time or over a specified period. What many mistakenly refer to as the incidence of congenital cardiac disease is in reality the prevalence, consisting of the number of newborns who are subsequently confirmed to have congenitally malformed hearts as observed within a defined population of live born individuals over a specified time. Such a proportion would more appropriately be called the prevalence at live birth. This prevalence at live birth reflects the incidence of congenital cardiac disease as modified by undetected conceptions, spontaneous and elective terminations, and stillbirths, all of which may occur differentially in those with specific aetiologies or defects. The recognised proportion can further be modified by the completeness and accuracy of ascertainment of disease in all those neonates born alive. Rarely, some estimates of prevalence may include stillbirths, giving the prevalence of congenital cardiac disease in completed pregnancies. Alternatively, some estimates may include fetuses assessed prenatally, some of which may be spontaneously or electively terminated. If it were possible accurately to assess disease prenatally in all fetuses from a uniform time point in gestation, then it would be possible to estimate the prevalence at fetal assessment. This is not the current reality, and the current inclusion of data relating to fetal diagnosis in estimates of prevalence creates great inaccuracy and confusion.
The prevalence at live birth is, of course, important in defining the maximal burden of congenital cardiac disease in the population. The natural and modified natural history specific to certain defects, and strategies for their management, influences the changing prevalence in the population over time, with the numerator and denominator both decreasing due to deaths, and the numerator falling due to spontaneous resolution. The absolute number and characteristics of patients at any given time is important in defining the burden posed by disease, which is a key determinant in defining the requirements for resources. With the increasing survival of patients into adulthood, and their transition into the system of health care providing for adults, this number has taken on increasing importance, but its accurate estimation is fraught with numerous methodological challenges.
RELEVANCE OF INCIDENCE AND PREVALENCE
Estimates of incidence and prevalence, and the study of factors influencing them, have relevance to many aspects of congenital cardiac disease and its management.
Aetiologic Associations
Accepting the caveats discussed above, it remains the case that knowledge of factors associated with variations in the prevalence of specific lesions, or groups of lesions, may suggest aetiologic influences. These factors might then be used to target prenatal screening and counseling, or they might be amenable to intervention, resulting in the prevention of congenital cardiac disease. In general, studies providing evidence of such associations have been difficult to perform when using an approach depending on cohorts identified at birth, since congenital cardiac disease is rare, specific lesions even rarer, and potential factors may similarly be rare. In addition, aetiologic factors may not act uniformly or consistently in respect to the lesion under investigation. Classifications of lesions on the basis of morphology may not relate to aetiology. Genetic studies have shown that specific genetic factors may be associated with a variety of lesions, which might not typically be grouped together on the basis of morphology. Studies of environmental factors have not yielded great discoveries, and have engendered few hypotheses. Focused hypotheses regarding associations between specific factors and specific lesions can more efficiently be pursued using studies based on case-control study designs.
Surveillance and Trends over Time
Related somewhat to aetiology, there is an ongoing interest in discovering trends in the prevalence of congenital cardiac disease. Ongoing surveillance is necessary to detect acute changes in prevalence that may be attributable to a specific aetiology, such as the epidemic of birth defects associated with the use of thalidomide during pregnancy. These changes in prevalence reflect a true change in the incidence of congenital cardiac disease. Many factors other than aetiologic ones, nonetheless, may contribute to a change in prevalence. Prevalence may decrease with prenatal diagnosis and elective termination. Prevalence may increase with increasing access to medical care, prenatal diagnosis, prevention of death during fetal life, improved screening and diagnostic capabilities, and their broader application. Estimates of prevalence may also change with alterations in definitions, nomenclature, and classification. Examination and comparison of prevalence across periods of time, or disparate geography and ethnic regions, requires careful attention to these details.
Likelihood of Diagnosis
Knowledge of the relative frequency of specific lesions can aid somewhat in formulating differential diagnoses and an index of suspicion. Most studies of prevalence at live birth continue to include cases that were diagnosed beyond the first few days of life, and often into the first year. In earlier studies, ascertainment relied almost exclusively on clinical manifestation and presentation. So-called critical disease is more likely to present in the neonatal period during the transition from fetal circulation with closure of the arterial duct and oval foramen, and the fall in pulmonary arterial resistance. Anomalies associated with evident syndromes are also more likely to be detected early. Some defects, such as ventricular septal defects, are not evident until further falls in pulmonary vascular resistance permit an increase in the shunting of blood, with emergence of heart murmurs and signs of cardiac failure, typically in early infancy. Still other defects, such as aortic valvar stenosis and coarctation, might initially be mild, without evident symptoms or signs, and not become manifest until the obstruction worsened weeks, months, or even years later. Some lesions, such as the aortic valve with two leaflets, or uncomplicated congenitally corrected transposition, may only present fortuitously many years later upon investigation for something else. This picture of diagnostic likelihood has changed recently. Some studies of prevalence have employed uniform early screening. Prenatal diagnosis has had an impact on age at presentation, independent of the severity or physiologic consequences of the cardiovascular defect. In addition, elective termination of more severe defects has had some impact on their relative frequency at live birth.
Burden of Disease and Requirements for Resources
Disease involving the cardiovascular system continues to be the greatest contributor to infant mortality related to congenital malformations. Trends, however, have shown significant decreases in mortality, and an increasing age at which death might occur. The greatest attributable factor to these trends has been advances in management, which have modified the natural history. To a lesser extent, decreasing prevalence of severe lesions related to prenatal diagnosis, and elective termination, may also have an impact. These factors increase the point prevalence of congenital cardiac disease at advancing ages, and have created a growing population of adults with congenitally malformed hearts, which has been predicted soon to exceed the point prevalence during childhood. There have been no studies which have specifically quantified the point prevalence beyond infancy or early childhood. These are difficult to attain, since they would require systems providing complete capture of data, and uniform ongoing contact with those providing health care. Studies in adults would be particularly problematic, in that broad-based systems for their health care do not yet exist, and many adults lose care through lack of coverage, and lack of available providers. In addition, many adults may falsely believe that their congenital cardiac malformation has been cured when it is increasingly shown to be associated with ongoing morbidity and risk of death. Current estimates of the prevalence of congenital cardiac disease in adults rely on projections and assumptions, rather than accurate and specific collection of data. Thus, beyond infancy, the true and ongoing prevalence is poorly known, and hence knowledge concerning the burden of disease on society is equally vague. In addition, while infants and children in general maintain their contact with those providing health care, and thereby influence the estimates made for requirements of resources and future planning in a well-defined system, this is not so for adults, for whom a well-defined system of care does not yet uniformly exist. Reliable estimates of prevalence in adults based on data are needed in order to plan for, and address, their needs for health care, as well as to assess their potential impact on the overall system providing care.
METHODOLOGY AND APPRAISAL
Estimates of prevalence can vary widely between different reports, depending on how the numerator was defined and ascertained, and the source of the data for the denominator. It can also be influenced by the period of time or point in time, and the absolute numbers in both the numerator and denominator. Ideally, differences in estimates of prevalence between reports should reflect true differences in prevalence related to potential aetiologic or epidemiologic factors, such as time trends or genetic or environmental differences between populations, rather than methodological discrepancies.
Denominator
When seeking to estimate prevalence, the first consideration should be the denominator, since that determines the population from which the numerator will be drawn, and also the perspective and utility of the estimate. The denominator is a key piece of information, since it defines the population to which the estimate may be applied, and suggests the population to which the estimate might be extrapolated. The most valid denominators are ones that are enumerated during the time of ascertainment of the cases, such as with a study designed on the basis of a prospective cohort. Such an approach gives the most valid and reliable value for the denominator, and also the numerator, but is intensive in terms of both time and resources. Most estimates of prevalence use, as their denominator, the total number of live births over a specified period of time derived from a geographically defined population, hence giving the prevalence at live birth. The defined population may truly be defined by geography, or may relate to a catchment area, or to a subset of the population under a specified plan or system providing health care.
A convenient, and readily available, denominator, frequently used, is the number of births reported to a governmental system for registration of births. These almost always relate to a geographically defined population, and reporting is usually mandatory. In looking at such estimates, it should be clear that the cases were ascertained from the same catchment or geographic area as the reported number of births from the birth registry. In addition, the periods over which the births occurred, and the cases were ascertained, should be identical. Information regarding the validity and reliability of the registry should also be available. Strategies by which all births were correctly identified, characterised, and reported should be explicit. Any information regarding validation, through audits of the data, should be sought. Some registries will include stillbirths and fetal deaths, and this information should be evident. Registries of births usually collect additional information, specifically demographics for the family, but may also include data regarding clinical diagnoses and characteristics at birth. Some estimates of prevalence will rely on this information for ascertainment of cases. If so, then the validity and reliability of that data needs carefully to be scrutinised.
Some studies report the prevalence of congenital cardiac disease as a contributor to overall or infant mortality. For these studies, a governmental registry of deaths may be used to define a denominator, that being the number of deaths occurring over a specified period of time within a defined geographic area. Many of the same limitations regarding validity and reliability of data for registration at birth apply to registration at death. Information regarding clinical diagnoses and cause of death are often collected. If this is used for ascertainment, then further consideration should be given to validity and reliability, particularly since the qualifications of the persons completing and submitting these data are typically unknown, as well as the definitions they might have used.
Numerator
The numerator of any estimate of prevalence will represent the number of cases of congenital cardiac disease that were identified from a population at risk over a defined period of time. To evaluate the numerator, one must know how the cases were identified and verified from the source of the data, what types of lesions were included and excluded, and what scheme was used for nomenclature and classification.
Ascertainment of Cases
The completeness of identification of the cases from the source population is a critical aspect of appraisal of the numerator of an estimate of prevalence. A comprehensive and active prospective surveillance of all sources of cases is likely to yield the most complete ascertainment, since ascertainment is a specific and planned endeavour. Most studies of this nature rely on clinical presentation, or evaluation of a living subject, as the initial point of entry. Many trivial lesions, nonetheless, such as small atrial septal defects, mild pulmonary valvar stenoses, tiny muscular ventricular septal defects, silent arterial ducts, and nonstenotic aortic valves with two leaflets, may not cause symptoms, and clinical findings may be subtle or absent. In addition, some of these lesions may resolve spontaneously without ever being detected. Still other types of severe lesions may occasionally lead to death before a clinical diagnosis is made or verified, and may not be ascertained unless autopsies are performed and the information is accessed and the case included. An important aspect of ascertainment is the interval from birth during which cases could be identified and included. This would be particularly important for cases identified initially on the basis of clinical features. The majority of patients with significant congenital cardiac disease would be expected to present early. Some lesions, however, may not become clinically manifest until the occurrence of physiologic changes, such as with a fall in pulmonary vascular resistance and increased shunting with ventricular septal defects, or with the development of pulmonary vascular disease or ventricular dysfunction with other lesions. Some lesions may become clinically manifest only with progression of the pathology over time, such as with valvar stenoses. The duration of follow-up for ascertainment must be sufficiently long that all cases are identified.
The optimal method for complete ascertainment would be a universal screening shortly after live birth with a technology with sufficient sensitivity to detect all lesions, usually echocardiography. Given the large number of newborns that would have to undergo screening to detect sufficient cases to allow such a valid and reliable estimate of prevalence, this is unlikely to be feasible. Currently, fetal ultrasound at 14 to 20 weeks of gestation seems to be universally performed for nearly all pregnancies in developed countries, particularly those with universal provisions for healthcare. This might afford the opportunity for a complete surveillance, but routine fetal ultrasound as performed has proven to have a low sensitivity, unless cardiac assessment is specifically and accurately applied. At present, fetal detection of cardiac lesions, even serious ones, is variable and imperfect. Other sources can be used to supplement cases ascertained from surveillance, including registries of deaths and autopsy records, registries of birth defects, and administrative data such as abstracts at discharge from hospital, and data from medical claims. These sources have the advantage that reporting is often mandatory, although they are limited by the validity and reliability with which diagnoses and details are reported.
Verification of Cases
Given the preceding discussion that most studies of prevalence rely on a clinical diagnosis for the majority of identification of cases, verification or confirmation is very important. Earlier studies focused on cases identified or verified by autopsy, but this would include only cases associated with a high mortality, and for whom a decision to perform an autopsy was made. Studies that have focused on cases ascertained through clinical presentation and identification have reported high estimates unless some sort of objective verification has been applied. Earlier studies have included cases verified by autopsy, surgery, or cardiac catheterisation. These estimates have been somewhat low, and skewed toward more serious lesions, but become inflated when cases with a clinical diagnosis only are included. The Baltimore–Washington Infant Study was one of the first studies to include echocardiography as a means of verification. 5 Echocardiography has since become the standard for the initial verification of cases, and has proven to have high sensitivity and specificity. In general, these early studies that used echocardiography as a means for verification have produced similar estimates to those studies which use more invasive means of verification, with larger estimates if cases with a clinical diagnosis only are also included. Broader and earlier application of echocardiography, and improved sensitivity through the application of cross sectional imaging and colour Doppler interrogation, have resulted in improved detection of haemodynamically insignificant lesions, such as tiny or small muscular ventricular septal defects, and silent arterial ducts. This improved ascertainment has largely accounted for an apparent increase in estimates of prevalence. The impact of fetal echocardiography will likely also influence ascertainment in future studies of prevalence, through both possible improved detection of cases with lesions who might succumb during fetal life, and through elective termination.
Sources of Data
Sources for the numerator vary widely in the degree to which ascertainment was passive or active, and the degree to which they accurately relate to a defined denominator. Knowledge of the details and limitations of these sources is important for critical appraisal.
Earlier studies tended to rely on registration of death and autopsy records for ascertainment. These studies were important in defining the spectrum of congenital cardiac disease and its morphologic features, particularly of complex lesions. Use of autopsy records tended to focus on cases presenting to a tertiary care centre, and were heavily weighted to more severe defects associated with a high mortality. The only valid denominator to which these cases might be applied is that of total autopsies at the given institution, and it would give the proportion of congenitally malformed hearts among deaths verified by autopsy, and may be an indicator of the contribution of congenital cardiac disease to overall mortality. Alternatively, most geopolitical regions have some sort of mandatory registration of death. There is usually an accurate enumeration of deaths, although reporting of causes of death and associated conditions may be less accurate and, therefore, less useful in determining a numerator. The accuracy of identification of cases from death certificates depends on who completes the registration, along with their knowledge of that person and their medical history, whether an autopsy was performed and the information included, and whether the congenital malformation was the primary cause of death, a contributing factor or an associated condition. One such study of death of patients with congenitally malformed hearts previously followed by paediatric cardiologists showed that only four-fifths of certificates noted cardiovascular disease, with the major defect being noted on only two-fifths of certificates. 6 This highlights the deficiencies in relying on registries of death for ascertainment.
Secondary data is that which was collected for purposes other than the research question at hand. Usually this data is collected for purposes of documenting clinical care, such as the health record, or for administrative purposes usually linked to reimbursement or statistical tracking. The findings often have the advantage that the collection itself is mandatory. Administrative data may include registries of births and deaths, and abstracts of discharge from hospital and databases of medical claims. The major limitation of using such data as the source of ascertainment is the variable completeness and accurateness by which diagnoses are recorded in a valid and reliable manner. Diagnoses recorded on registrations of births and deaths may only enumerate those present and verified at the time of registration. Ascertainment using databases of hospital discharges only identify those cases associated with hospitalisation, which might be true for only a limited number of lesions. 7 The addition of databases recording medical claims that capture outpatient assessments and diagnostic testing may be important supplements to ascertainment using administrative data. While administrative data is of limited value in specifying the numerator, as mentioned previously, it is often useful in defining the denominator, particularly of live births.
An important source of secondary data for ascertainment is the medical record. Sufficient information can often be found in the medical record to identify and verify cases. Many studies have used the medical records of tertiary care centres in such a manner in reported series, but are limited in that the denominator is difficult to define. The challenge of the denominator often precludes an accurate estimate of prevalence, but these studies are useful in defining the spectrum and distribution of congenitally malformed hearts, particularly in reference to the provision of clinical care. These types of studies are now rarely reported, except from developing countries where the only cases that may be enumerated are those presenting to a specialised centre. Some studies relate to a defined system for provision of healthcare, with the denominator being all individuals captured and registered within that system. These studies can produce an estimate of prevalence, but the degree to which it is representative of the proportion of the population not included within that system is often unknown.
The most accurate and complete source for ascertainment is from surveillance, particularly if the surveillance is an active rather than passive process. Mechanisms for passive surveillance tend to rely on voluntary reporting, which may be incomplete with variable verification of diagnoses. The most valid estimates of prevalence are derived from active surveillance, particularly if both the numerator and denominator are defined simultaneously as part of the surveillance. Such active studies tend to be designed specifically to determine an estimate of prevalence. They may also include a more focused collection of data, usually in the format of case-control design, aimed at detecting associations of a potential aetiologic nature. The completeness of ascertainment depends on the comprehensive nature of the system created for identification. This may include primary care networks, paediatric cardiologists working in both community-based and tertiary care practise, as well as all points of obstetrical delivery and care of the newborn. The reporting from these points of contact may be voluntary by the practitioners themselves, or may involve surveyors deployed to these sites. The reporting may occur as each case is identified, or after periodic assessments of logs or records. Identified cases are usually referred to experts for verification and specification, usually in centres of tertiary care. Pathology departments may also be involved. Additionally, registrations of births and deaths, and other types of administrative data, may be reviewed. The challenge to complete ascertainment is the reliance on clinical presentation and findings to initiate identification and verification of the cases, the limitations of which have been stated above.
The most complete source for ascertainment would be the active screening of a defined population. A sensitive and accurate method of identification would be universally applied to all subjects included in the denominator. Clinical assessment by paediatric cardiologists has been shown to have sufficient, though not perfect, sensitivity and specificity, while other types of providers have fared less well. 8,9 Verification and specification of clinically suspected cases with echocardiography could then be applied. Alternatively, echocardiography could be universally applied as the screening tool, ensuring complete and accurate ascertainment. These types of screenings are not feasible for application on a large scale, and the resulting estimates of prevalence would tend to be more accurate but less reliable.
Definition of Cases
Regardless of the source used for ascertainment, it must be clearly stated which specific lesions were included and excluded from the numerator. This predominately applies to lesions of minor or no functional consequence but which are common, as discussed previously. The inclusion or exclusion of these types of lesions can have an important impact on inflating the prevalence estimate. Additionally, the time period after live birth during which cases could be identified should be specified and reported. It should be clearly stated whether cases noted in stillbirths and cases diagnosed by fetal assessment, but either spontaneously or electively aborted, are included or excluded.
Nomenclature and Classification
In addition to specifying an overall prevalence, most studies provide a breakdown by specific lesions. This is challenging, in that some lesions may exist in isolation, such as partially anomalous pulmonary venous connection, or as part of a complex of lesions, such as anomalous pulmonary venous return in the setting of isomerism of the atrial appendages and complex congenital cardiac disease. Some isolated lesions may also be complicated by additional lesions, such as transposition with an associated ventricular septal defect. Some patients may have two lesions of equal importance, such as a ventricular septal defect and coarctation of the aorta. Additional lesions may be acquired as either part of the natural history, or as a result of therapeutic interventions. These distinctions make it challenging to identify primary as opposed to secondary lesions. These challenges are magnified by the lack of a uniform nomenclature. Lesions had initially been described on the basis of their first reports, such as Ebstein’s malformation or the Holmes heart, and then from systematic examination of pathologic specimens. Nomenclature was therefore driven by considerations of morphology. Increasing understanding of both physiology and cardiac development have modified some of these descriptions, and have led to some controversies. As a result, individuals have used systems for description that were developed within the institutions where they had either trained or practised. Different organisations have attempted to develop consensus regarding common nomenclature, and recently have come together as the International Nomenclature Working Group. 2,3 This organisation seeks to cross-map the different systems, and to provide a common, consensus-based system. 4
There is often a desire to group together lesions into categories that reflect a common basis in either development, morphology, physiology, management, outcome, or aetiology. A scheme aimed at both the presumed developmental mechanism and physiology was employed in the Baltimore–Washington Infant Study (Table 3.2 in their monograph). 5 It was hypothesised that this classification might reduce aetiologic heterogeneity when related to potential familial and environmental risk factors. More recent studies of genetic mechanisms have shown that a single genetic defect may be associated with a more diverse range of morphologic abnormalities, and that these may not be concordant with the scheme based on developmental mechanisms. These genetic studies have refined our understanding of mechanisms of cardiac development that may influence future schemes for classification.
Reliability of Estimates
When reporting an estimate, it is important to be clear about the methodology behind determining the numerator and the denominator. When describing an estimate, details are needed of what the number represents. Prevalence at live birth is usually reported in terms of the number of cases of congenital cardiac disease in each 1000 live births. Additional information may include the geographic location, and the years over which the numerator and denominator were derived. Not all estimates made in this fashion have the same reliability. Reliability is influenced by the duration underlying the estimate, and the absolute magnitude of the numerator and denominator. Studies in which the estimate represents a period of many years are less reliable, in that changes are likely to have occurred in the system, and influenced the estimate. The absolute magnitude of the numerator and denominator influence the width of the confidence limit. A confidence limit is a statistical derivation of an interval in which we may be 95% confident, or sometimes 70% confident, that the estimate is true. The width of the interval is highly dependent on the absolute numbers. The greater the numbers, the narrower the width of the interval, and the more reliable the estimate. Confidence limits should be reported with any estimate, and often the absolute numbers for the numerator and denominator should be given. The absolute and relative magnitude of the numerator in relation to the denominator may also influence reliability. Estimates based on small numbers of cases are less reliable, and the confidence limits enclose wider ranges of relative magnitude. For example, a 95% confidence interval around an estimate of 100 per 1000 live births might be reported as 98 to 102, or ±2% of 100. A different study of similar size noting an estimate of 10 per 1000 live births with a similar 95% confidence interval of 8 to 12 in terms of absolute magnitude, but ±20% of 10 in terms of relative magnitude, thus would be less reliable. Thus, estimates for total prevalence of congenital cardiac disease, and for common specific lesions, tend to be more precise and reliable than those for less common specific lesions. In addition to reporting confidence intervals around estimates of prevalence, such intervals should also be reported around any measure of association, such as relative risk, odds ratios, or attributable risk.
Critical Appraisal
Most studies that report estimates of prevalence are specifically designed to do so, and have it as the primary aim of the study. Before accepting the results of these studies, it is important critically to appraise the methods and reporting, in order to determine if the findings are valid, reliable, and relevant. Critical appraisal is also required when wishing to compare estimates from different studies, such as comparisons of trends over time, or different populations. Consideration must be given to all of the aspects of the denominator, numerator, and reporting already described. In Table 8-1 , I have delineated questions for which the answers should be evident when critically appraising a report regarding prevalence.
| What was the stated purpose or aim of the study? |
| How accurate and valid was the estimate reported? |
|
|
| How reliable is the estimate of prevalence estimate? Are confidence intervals provided? |
| How does the estimate from the study compare to those reported from other studies with comparable methodology? |
| Is the estimate applicable to your own clinical population? Is the population studied similar to your own clinical population in terms of setting, time, geography, and demographic characteristics? Is the estimate of prevalence relevant to your own clinical or research question? |
FACTORS INFLUENCING ESTIMATES OF PREVALENCE
Ideally, factors should be identified that might have a true causal relationship to the development of congenital cardiac disease, and hence influence the true incidence. Other factors may alter outcomes during fetal life, and predominately affect the prevalence at live birth. Identification of these factors may allow for the prevention of congenital cardiac disease, or provide new knowledge as to aetiology and development. The identification of these associations comes mainly from observational studies.
Environmental Factors
The most commonly held belief is that congenital cardiac disease is the product of an interaction between genes and their environment. Evidence, however, has yet to progress sufficiently far to prove this notion. Numerous environmental factors have been linked to such development, independent of any already known genetic influence or predisposition. The Baltimore–Washington Infant Study provided a thorough assessment of environmental and parental risk factors for the development of congenital cardiac disease, and development of specific lesions. 5 Other research has since contributed to this growing body of knowledge. It is important to note that association does not infer causality, and that identified risk factors can be viewed only as potential risk factors. To infer causality, it would be necessary to perform a randomised controlled trial, and impose a controlled exposure on mothers with child. The ethical dilemma in asking expectant mothers intentionally to expose themselves and their unborn child to a potentially harmful agent for the purposes of a randomised controlled trial can be imagined. We must rely, therefore, on observational studies. Initial associations are usually determined from studies of cohort, where exposures are determined by recall after the child has been born and identified as having congenital cardiac disease. These studies are often subject to recall bias. Case-control studies are an efficient method to focus on a specific exposure, comparing the prevalence of congenital cardiac disease in those having a specific exposure to those not confronted by the same exposure. This is usually reported as an odds ratio, being the odds of congenital cardiac disease in those with the exposure divided by the odds of the disease in those without the exposure.
A recent Scientific Statement from the American Heart Association reviewed studies of prevalence, and summarised studies of environmental factors. 10 In Table 8-2 , positive associations for maternal illnesses and maternal exposures are highlighted.
| Exposures Associated with Definite or Possible Risk of Offspring with Any Congenital Cardiovascular Defect | Odds Ratio |
|---|---|
| Maternal Illness | |
| Phenylketonuria 11–14 | >6 |
| Pregestational diabetes 15–18 | 3.1–18 |
| Febrile illness 15, 19–21 | 1.8–2.9 |
| Influenza 21, 22 | 2.1 |
| Maternal rubella 23 | † |
| Epilepsy 24 | † |
| Prepregnancy overweight/obesity 25 | 1.13–1.40 |
| Maternal Therapeutic Drug Exposure | |
| Anticonvulsants 26, 27 | 4.2 |
| Ibuprofen 28 | 1.86 |
| Sulfasalazine 29 | 3.4 |
| Thalidomide 30 | † |
| Trimethoprim-sulfonamide ∗ ,29, 31 | 2.1–4.8 |
| Vitamin A congeners/retinoids ∗ ,32, 33 | † |
∗ Risk reduced if mother took folic acid simultaneously.
Table 8-3 shows that maternal exposure to organic solvents has frequently and consistently been reported to be associated with relatively high odds ratios for development of specific congenital cardiac malformations. In contrast to maternal illnesses and use of medications, it is feasible to avoid exposure to organic solvents, and programmes can be directed at reducing maternal exposures to these materials.
| Defect | Risk Ratio |
|---|---|
| Defects of the ventricular outflow tracts 34, 35 | 2.0–3.9 |
| Hypoplastic left heart syndrome 15 | 3.4 |
| Coarctation of the aorta 15, 36 | 3.2 |
| Pulmonary stenosis 15 | 5.0 |
| Transposition 15 | 3.4 |
| Tetralogy of Fallot 15 | 2.7 |
| Totally anomalous pulmonary venous return 15, 37 | 2.0 |
| Atrioventricular septal defect (nonchromosomal) 15 | 5.6 |
| Ebstein’s malformation 15, 38 | 3.6 |
An earlier study summarised environmental associations noted from the Baltimore–Washington Infant Study. 2 The study was a large population-based surveillance with careful verification. The investigators reported results as both relative risk, representing the prevalence of congenital cardiac disease in those with the exposure divided by the prevalence in those without the exposure, and attributable fraction, this being the proportion of cases of specific defects that might be attributable to specific exposure. In Table 8-4 , the identified positive associations are highlighted. Some of the significant exposures were paternal rather than maternal.
| Malformation and Potential Risk Factors | P < 0.01 | Relative Risk | |
|---|---|---|---|
| AF (%) | 95% CI | ||
| Transposition with Intact Ventricular Septum (N = 106) | 12.1 | 8.5–15.8 | |
| Influenza | 7.0 | 3.6–10.3 | 2.2 |
| Miscellaneous solvents | 4.8 | 3.0–6.6 | 3.2 |
| Tetralogy of Fallot (N = 204) | 6.5 | 4.8–8.3 | |
| Paternal anaesthesia | 3.9 | 2.4–5.5 | 2.5 |
| Clomiphene | 2.4 | 1.5–3.4 | 3.0 |
| Atrioventricular Septal Defect with Down Syndrome (N = 190) | 4.6 | 2.7–6.5 | |
| Ibuprofen | 4.6 | 2.7–6.5 | 2.4 |
| Hypoplastic Left Heart Syndrome (N = 138) | 8.6 | 6.9–10.3 | |
| Solvent/degreasing agent | 4.6 | 3.2–6.0 | 3.4 |
| Family history of congenital heart disease | 4.0 | 3.1–4.9 | 4.8 |
| Coarctation of the Aorta (N = 120) | 9.4 | 8.1–10.8 | |
| Family history of congenital heart disease | 4.6 | 3.5–5.7 | 4.6 |
| Macrodantin | 2.3 | 1.8–2.8 | 6.7 |
| Clomiphene | 2.0 | 1.4–2.7 | 4.5 |
| Isolated/Simple Perimembranous VSD (N = 459) | 7.9 | 4.2–11.6 | |
| Paternal use of marijuana | 6.0 | 2.2–9.7 | 1.4 |
| Maternal use of cocaine | 1.7 | 0.9–2.5 | 2.4 |
| Multiple/Multiplex Perimembranous VSD (N = 181) | 8.3 | 6.0–10.5 | |
| Paternal use of cocaine | 4.8 | 2.6–6.9 | 2.3 |
| Diabetes mellitus | 2.1 | 1.4–2.8 | 3.9 |
| Metronidazole | 1.4 | 1.1–1.7 | 7.6 |
| Atrial Septal Defect (N = 187) | 14.1 | 11.3–17.0 | |
| Gestational diabetes mellitus | 4.4 | 2.5–6.2 | 2.4 |
| Paternal use of cocaine | 3.7 | 1.9–5.4 | 2.3 |
| Family history of congenital heart disease | 3.4 | 2.4–4.3 | 3.9 |
| Corticosteroids | 2.6 | 1.9–3.2 | 4.8 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
