SESSION II. PUBLIC HEALTH ASSESSMENT OF PI DISEASES
Applying Genetics and Public Health Strategies to Primary Immunodeficiency Diseases
November 8-9, 2001 ~ Atlanta, Georgia
Prepared by: Office of Genomics and Disease Prevention, Centers for Disease Control and Prevention Department of Health and Human Services
SESSION II. PUBLIC HEALTH ASSESSMENT OF PI DISEASES
Disease registries are designed to actively collect patient-specific information from multiple sources over time. They can be either case or hospital-based or population-based. Case-based registries are usually designed to improve patient care but can be useful for studying rare diseases. Case-based registries provide useful clinical and research information but cannot be used to calculate population-based rates; limitations include lack of representativeness and standardized case definitions. Population-based registries collect standardized information on all cases of a disease in a geographically defined area. Data can be used to determine disease incidence, identify subgroups at increased risk, examine trends, and evaluate the effectiveness of screening and treatment programs.
Dr. Stiehm summarized information about the prevalence and distribution of PI diseases as estimated from registry data and a survey conducted by the Immune Deficiency Foundation (IDF). Prevalence estimates for PI diseases from three published registries in Australia (Baumgart et al. J Allergy Clin Immunol 1997), Switzerland (Affentranger et al. Immunodeficiency 1993), and Norway (Stray-Pedersen et al. J Clin Immunol 2000) are 1:36,000, 1:14,000, and 1:14,000, respectively. An IDF survey conducted in 1995 identified 25,000 patients with primary immunodeficiency and added another 25,000 as an estimated number of unidentified patients. This provided an estimated prevalence in the United States of 1:5,000 (50,000/250 million population). The distribution included CVID (5,300 patients), IgA deficiency (5,237), IgG subclass deficiencies (4,973), and other genetically defined immunodeficiencies (5,000). Of the total number of patients, 40 percent were <18 years of age, 52 percent were female, and 70 percent reported being treated with intravenous gamma globulin (IVIG).
Additional distribution estimates are derived from the work of Dr. Stiehm and colleagues, who combined four recent registries (Spain, Australia, Latin America, and Norway) with a total of 3,369 patients to estimate the relative frequency of various PI disorders. In this analysis, 14 disorders represented 90 percent of cases. Antibody defects comprised 62 percent of cases; combined defects, 13 percent; phagocytic defects, 6 percent; cellular defects, 5 percent; and complement defects, 4 percent. The most common defects were IgA deficiency (925 patients), CVID (614 patients), ataxia-telangiectasia (205 patients), XLA (214 patients), IgG subclass deficiencies (165 patients), and SCID (115 patients). About 60 percent of patients were male, and more than half were adults.
Dr. Stiehm concluded that the overall prevalence of PI diseases of all types might be as high as 1:5,000. Many late-onset cases are diagnosed as CVID, IgA deficiency, and IgG subclass deficiency. All registries include more adults than children. Quantitative neutrophil defects, hyper-IgE syndrome, and DiGeorge syndrome are likely underreported.
In 1992, the Immune Deficiency Foundation received a contract from NIH to establish and maintain a registry of U.S. residents affected by several PI diseases. The goals were to estimate minimum incidence, characterize epidemiologic and clinical features, and facilitate research. The project was initiated with a registry for patients with CGD and was expanded in 1997 to add seven other diseases: hyper-IgM syndrome, XLA, CVID, WAS, SCID, LAD, and DiGeorge syndrome.
In 1996, a survey was sent to 17,400 members of seven professional societies and to chairpersons of departments of pediatrics and medicine. Of 3,680 physicians who responded, 1,108 reported 15,927 patients. Consenting physicians were sent a clinical data entry form requesting patient information about demographics, diagnosis, family history, clinical course, and treatment. Data were entered into a relational database. Confidentiality was ensured because only physicians were contacted and patients were de-identified.
A main goal is to provide a minimal estimate of the frequency of the disorders in the United States. The most mature data are for CGD; the minimum estimate of birth rate is 1:200,000. Birth rates for the other disorders (e.g., 1:600,000 for WAS; 1:1 million for XLA) are probably underestimates. The registry also is being used to develop clinical descriptions of each disorder. For example, the autosomal recessive type of CGD is milder than the X-linked recessive disease. Patients with X-linked disease are significantly younger at diagnosis than those with autosomal recessive disorder and have a higher prevalence of certain infections and chronic inflammatory disease manifestations.
The registry is a useful resource for collaborative research and improved patient access to new or novel therapies. Completed projects using registry data include studies on host defense mechanisms in CGD, stem cell transplantation in CGD, and hepatitis C and enteroviral infections in PI diseases. Ongoing research is addressing such topics as malignancies in WAS and racial background of PI patients. Advantages of the registry center on the integrity of the data and the availability of validated diagnoses, previously unavailable clinical and epidemiologic data, and a national resource for research. Limitations include potential ascertainment bias, dependence on physician cooperation, lack of access to patients, and a tension between the availability of too little information from the registry’s point of view and the need for too much information from the physician’s point of view.
Dr. Smith reported on the registry of PI diseases established by the European Society for Immunodeficiencies (ESID) in 1994. The European registry includes diagnostic information, immunoglobulin values, therapies, and family histories. Physicians submit data either to their national registries, if available, or to both national and ESID registries. National registries submit information to the central ESID registry at Karolinska Institutet, Sweden. Data are submitted on one form that contains all pertinent diagnostic information and facilitates computerized data entry and retrieval. All communication is through physicians; patients cannot be contacted directly.
As of July 2000, the registry included clinical data for 8,903 patients representing 26 countries; more information can be obtained at www.esid.orgexternal icon. The continued growth of the registry is due mainly to the submission of individual patient reports from physicians in ESID countries. So far, patients with primary antibody deficiencies comprise the majority of entries, although reports from member countries vary considerably. Among antibody deficiencies, the distribution is as follows: IgA deficiency (2,582 patients), CVID (1,764 patients), IgG subclass deficiency (684), XLA (504), and hyper-IgM syndrome (140).
In 1995, an international study group established the first locus-specific, web-based immunodeficiency mutation database. Designated BTKbase and initially maintained at Karolinska Institutet, the database collects information about mutations in the BTK gene, which cause XLA. Similar locus-specific databases have been initiated for other inherited immunodeficiencies using the BTKbase model and are maintained at the University of Tampere, Finland. The main registry and the locus-specific databases are funded mainly through the European Union. Mutation data can be submitted directly to the database coordinators or through web-based methods. Information from the databases can be used to analyze the distribution of mutation types and distribution in exons and introns, including mutational hotspots, as well as their location in protein domains. Consequences of mutations to the encoded proteins also have been evaluated.
Discussion–A brief discussion generated these points:
- Ideally, registry data should be more accessible and data analysis more efficient, but patient integrity cannot be compromised.
- Clear diagnostic criteria and case definitions are essential for deriving useful assessment data from registries. Maintaining up-to-date
patient information is a major undertaking.
- Making current registries more population-based would be an enormous, resource-intensive undertaking. In Europe, registries will be nation-based for the foreseeable future. The U.S. registry hopes to develop a sampling frame that is more population-based, e.g., through third-party payers, physician ZIP codes, or professional societies.
- Deriving incidence rates for PI diseases requires population-based screening. Patients may die of infections before the infections are recognized. The key is to find undiagnosed patients.
The next presentations addressed the potential for the use of existing population-based data and data-collection models to assess the prevalence of PI diseases and associated health outcomes.
Several existing databases available through federal, state, and other sources may be useful in studying indicators of the impact of PI diseases. Many databases are publicly available and are either free or of minimal cost. Three types of data sources might be helpful for the study of PI diseases: hospital-discharge databases, large linked databases from managed-care organizations, and national death certificate data.
Hospital-discharge databases–The Health Care Cost and Utilization Project (HCUP) is a federal-state-industry partnership to develop standardized databases for health services research and health policy analysis. HCUP is maintained by the Agency for Health Care Research and Quality (AHRQ), which develops HCUP databases and makes them available for restricted public access. The data are organized in two ways–either as the Nationwide Inpatient Sample, a 20 percent sample of hospitalizations, or as the State Inpatient Database, which includes data on all hospitalizations in participating states. The Nationwide Inpatient Sample includes data on approximately 6 million inpatient stays from a national sample of approximately 900 hospitals. The State Inpatient Database covers inpatient care in community hospitals in participating states, representing more than half of U.S. hospital discharges. The uniform data in HCUP facilitate comparative studies of health care services, variations in medical practice, effectiveness of treatments, and use of services by special populations. As an example of the usefulness of hospital discharge data, Yoon, et al. (Arch Pediatr Adolesc Med 1997) used 1991 data from two states to document the substantial morbidity rates and hospitalization charges associated with birth defects and genetic diseases in the pediatric population.
The National Hospital Discharge Survey, sponsored by CDC’s National Center for Health Statistics (NCHS), is another source of hospital discharge data. Most hospital discharge data sets have one; typically, they enumerate hospital discharges rather than individual patients, thus limiting their usefulness for certain studies.
Large, linked databases from managed-care organizations–The Vaccine Safety Datalink (VSD) is a partnership between CDC and four health maintenance organizations to evaluate vaccine safety in children in a large-scale prospective study. Computerized data on vaccination, medical outcomes, and health services use are provided for a well-defined population of approximately one million children (1993-1996). Data on hospitalizations and emergency department and clinic visits are also included. Because of the large number of enrollees, large populations can be examined for relatively infrequent events. For example, Belay, et al. (Pediatr Infect Dis J 2000) used VSD data to determine the incidence of Kawasaki syndrome and identify differences related to sex.
National death certificate data–Since 1968, NCHS has compiled data from all death certificates filed in the United States and has made these data available in its Multiple-Cause Mortality Files. The files include demographic and geographic information about the decedent and International Classification of Disease (ICD) codes for the underlying cause of death and up to 20 conditions listed on the death certificate. Dr. Rasmussen and colleagues used Multiple-Cause Mortality Files for 1983-1997 to study deaths associated with neurofibromatosis 1 (NF1). They identified 3,770 cases of presumed NF1 among approximately 32 million deaths, for a prevalence of 1:8,700. Mean and median ages at death were 54.4 and 59 years, respectively, compared with 70.1 and 74 years in the general population. Malignant neoplasms also occurred more frequently than expected among persons with NF1. The use of mortality files for this study allowed a population-based analysis and comparison with data from the general population.
An obstacle to the analysis of PI diseases using Multiple-Cause Mortality Files is the lack of specific and unique codes for disorders of the immune system. A preliminary analysis of mortality data using ICD-9 code 279.1 (deficiency of cell-mediated immunity) for 1987-1998 yielded 382 deaths among approximately 26.8 million deaths (excluding codes for HIV infection), or a rate of 1:70,000. Most of these cases were among children under 10 years of age, consistent with deficiencies of cell-mediated immunity. The longer survival time of some cases, however, raised concerns about the inclusion of other conditions. ICD-9-CM (clinical modification) codes are more promising. The recently released ICD-10 provides more specificity and has been used to code mortality data since 1999; ICD-10-CM is being developed. Dr. Rasmussen concluded that public databases have several strengths: they are population-based, include large numbers of cases, collect data on cases and controls, and are relatively inexpensive. Limitations center on the quality of initial data collection, biases in ascertainment of disease, and the inability of some databases to distinguish hospitalizations from cases; anonymity may also limit the potential for linkages to other data. These databases can be valuable if investigators recognize the limitations, try to improve the quality of the original data measurements, and address biases. Recommendations are to explore the use of these databases for the study of single-gene disorders, explore the impact of coding changes
(ICD-10; ICD-10-CM), and participate in ICD revisions to promote the development of unique and specific codes.
Dr. Edmonds reported on the goals and methods of ongoing public health surveillance, using birth defects surveillance as an example. Birth defects are the leading cause of infant mortality and contribute substantially to illness and long-term disability. Each year, 120,000 to 160,000 children are born with major birth defects, and children with birth defects account for 30 percent of admissions to pediatric hospitals. In 1992, the 15 most significant birth defects cost the nation $8 billion. Given the public health importance of birth defects, surveillance systems are needed to measure morbidity and mortality, monitor trends, stimulate research, provide data for etiologic studies, evaluate the need for and facilitate access to services, guide and assess the progress of prevention interventions, and inform education and advocacy efforts. These purposes are best accomplished through surveillance systems that use multiple data sources, have accurate and precise diagnostic criteria, perform timely data analysis and dissemination, use personal identifiers for follow-up and data linkage, and guarantee confidentiality.
Birth defects surveillance began in part as a public health response to the thalidomide-related adverse events of the late 1950s and early 1960s. Birth defects data sources include vital records, hospital and clinic records, administrative databases, prenatal diagnosis centers, and clinical examination findings. Case ascertainment methods yield varying rates of birth defects, ranging from 1.5 percent for review of birth certificate data and 3.4 percent for mandatory hospital reporting to 8.3 percent for examination of every baby born.
Accuracy of coding is a central challenge. The ideal is for codes to be specific and unique enough to generate a one-to-one correspondence between the features of a child with birth defects and a set of codes. In practice, however, relatively few syndromes have specific ICD codes, and existing codes are not sufficiently detailed for birth defects surveillance. Additional challenges include timeliness of reporting, consistency of diagnoses, balance between quality and timeliness of data, confidentiality, ability to share data, and duplication of cases.
One example of a population-based birth defects surveillance system is the CDC-sponsored Metropolitan Atlanta Congenital Defects Program (MACDP), which has been in continuous operation since 1968. Through intensive case ascertainment of major birth defects, MACDP serves as a case registry for epidemiologic studies, a prototype for other birth defects surveillance systems, and a laboratory for testing new surveillance methodologies. CDC has also awarded cooperative agreements to 36 states to enhance state-based birth defects surveillance, facilitate data sharing, and add to knowledge about rare birth defects and geographic variation.
In 1996, Congress enacted legislation directing CDC to establish the Centers for Birth Defects Research and Prevention. Eight Centers participate in the National Birth Defect Prevention Study (a collaborative case-control study of infants with major birth defects), conduct genetic and environmental epidemiologic studies, and enhance state surveillance systems. Each state approaches birth defects surveillance differently. The diversity of approaches–particularly methodologies used to generate timely data, applications to monitor prevention activities, and projects to improve access to health services and early intervention–provides a useful resource for development of surveillance systems for other childhood diseases. Planning for new surveillance systems should include interagency coordination; involvement of an advisory committee; development of goals, objectives, case definitions, and methodology; assessment of needs and resources, and awareness of relevant legislation.
Dr. Campbell gave the first of two presentations on efforts to collect epidemiologic and surveillance data on patients with other genetic diseases, which may be useful models for the assessment of PI diseases. The Cystic Fibrosis Foundation (CFF) supports and accredits 117 CF care centers nationwide. These centers provide a national network of specialized care for persons with CF, offer comprehensive diagnosis and treatment, and participate in clinical trials. Since 1966, CFF has sponsored the National Cystic Fibrosis Patient Registry, requiring
CFF-accredited care centers to complete standardized questionnaires for all patients. Data are entered electronically at the centers and submitted annually to CFF. Fixed costs total approximately $200,00 per year, but additional payments are made to participating centers to help support the multidisciplinary teams who see patients and collect the data.
The ultimate goal of the registry is to improve survival while maintaining patient privacy. Strengths center on its effectiveness as an epidemiologic tool and a mechanism to identify patients for clinical trials or new therapies. Weaknesses include data entry errors and duplication. In its current form, the registry is not a patient management tool, and it has limited ability to support outcomes research. Data entry and management, even for the limited information collected, are time consuming and resource intensive.
The registry has been successful in meeting its main goal: survival of CF patients has tripled since its initiation. Registry data are used to conduct epidemiologic studies, direct future research, and design clinical trials. A comprehensive annual report summarizes data in graphic form on all patients under care at CF care centers in the United States.
CFF also publishes Clinical Practice Guidelines for Cystic Fibrosis, a summary of current evidence and expert opinion that sets national standards of care. The guidelines are developed independently from the registry and do not necessarily reflect registry data. A goal is to merge the two activities to close the gap between standards (guidelines) and practice (registry) by addressing the national variability in treatment and outcomes, incorporating data and lessons learned from CF care centers, and ensuring consistent implementation of practice guidelines. Methods for accelerating improvements in clinical care are to identify high-leverage areas for clinical improvement and measure them using the registry, update the Clinical Practice Guidelines, and move data from the center level to the patient level. Future changes to comply with Health Insurance Portability and Accountability Act (HIPPA) standards are to add informed consent, move to a web-based system, and provide patient access to data. Three pilot studies are being conducted to test encounter-based data entry.
As mandated by Congress, the mission of CDC’s Hematologic Disease Branch, Division of AIDS, STD, and TB Laboratory Research, NCID, is to reduce or prevent complications of hemophilia and other bleeding and clotting disorders. In response to the mandate, CDC established a comprehensive program that includes a national surveillance system, resources for prevention interventions through hemophilia treatment centers (HTCs) and peer-led organizations, and epidemiologic and prevention research.
CDC’s first surveillance effort, conducted through cooperative agreements with six state health departments, was designed to (1) identify all patients with hemophilia in the states; (2) characterize the population in terms of disease type, mortality, resource use, and sources of care, complications, and joint disease; and (3) identify risk factors and outcomes of care through medical chart abstraction. Through this effort, the first population-based estimate of hemophilia prevalence in the United States was established, and the effectiveness of the comprehensive care model implemented by HTCs was demonstrated. The method provided generalizable data and established a retrospective cohort for further study but was costly and labor-intensive. Other limitations included the inability to contact patients, obtain information not provided in medical charts, collect specimens, or validate tests. Human subject confidentiality policies now in place will limit access to data sources.
In 1996, CDC initiated the Universal Data Collection surveillance system in collaboration with 140 hospital-based HTCs to (1) monitor blood safety among recipients of blood products and develop a specimen repository, (2) monitor the extent and progression of joint disease, and (3) identify issues for further study. All patients receiving comprehensive care through HTCs are asked to participate, and routine clinical data and a blood specimen are obtained upon consent. The surveillance system is prospective and will increase in value over time. It will also guide clinical practice and establish standardized measurements and testing procedures. The burden on clinicians is minimal, and additional modules can be added. Limitations include the need for Institutional Review Board (IRB) approval at every study site, limitations on the amount of data collected, inability to collect information on patients outside the HTC network, and dependence on infrastructure support and active efforts to ensure high-quality data.
Discussion–The participants made these points:
- One option for surveillance of PI diseases is to piggyback onto CDC’s birth defects surveillance activities. The infrastructure is already in place, and the birth defects surveillance system collects DNA samples.
- Minimal resources for data collection can be maximized by linking registries that are already surveying large populations of children. Parents and providers favor linked registries and integrated data sets.
- Immunization registries are evolving into children’s health registries. Some states have had some success in merging registries.