- Email: [email protected]
Estimates of birth prevalence rates of spina bifida in the United States from computer-generated maps
Estimates of birth prevalence rates of spina bifida in the United States from computer-generated maps FRANK
GREENBERG,
LEVY
M. JAMES,
GODFREY Atlanta,
M.D. M.S.
P. OAKLEY,
JR.,
M.D.,
M.S.P.M.
Georgia
A computer-generated mapping procedure was developed to estimate geographic and race-specific birth prevalence rates of open spina bifida. The estimates are based on birth certificate data adjusted for underascertainment. Separate maps were produced for white births and black births. For both races there is a general decreasing rate of spina biida from east to west. The highest rates are eight per 10,000 total births for whites in Southern Appalachia, and the lowest rates are less than one per 10,000 for blacks in the Rocky Mountain states and the Pacific Northwest. Until more exact data are available, these maps represent the best current available data on racial and geographic birth prevalence rates in the United States. They are useful for program planning and as an aid in interpreting maternal serum a-fetoprotein levels to detect neural tube defects. (AM. J. OBSTET.
GYNECOL.
145:570,
1983.)
THE NEURALTUBE DEFECTS anencephaly and spina bifida are among the most common and serious birth defects in the United States. The estimated birth prevalence rate of neural tube defects (the number of cases of neural tube defects at birth in a given area divided by the total number of live births and stillbirths in the area) in the early 1970s was 16 per 10,000 total births in the United States as a whole.’ However, there are known differences in rates when data are analyzed by race, geographic region, and time. The rate of neural tube defects among white births is two to three times higher than that among black births.2 Rates are higher in the eastern and southern United States than in the western United States, and there has been a decreasing secular trend in rates in both races and in all geographic regions.3 Recently, there has been an increasing need to provide more specific birth prevalence data on neural tube
From the Birth Defects Brunch, Chronic Diseases Division, Center for Environmental Health, Centers for Disease Control, Public Health S&e, United States Department of Health and Human Sentices. Received
for publication
Accepted
September
Reprint Branch, Control,
570
requests: Chronic Atlanta,
May
11, 1982.
2, 1982. G. P. Oakley, M.D., Birth Defecctr Diseases Division, Centers for Disease Georgia 30333.
defects, primarily spina bifida, with particular regard to racial and geographic distributions. Spina bifida is generally considered to be of greater concern because it is a major cause of morbidity from physical and mental handicaps among infants and children.“ Data on birth prevalence of spina bifida are useful for several reasons. First, they can be used for program planning for the care and treatment of affected infants and children. Second, they can be helpful in the decision to initiate a local maternal serum a-fetoprotein (AFP) screening program to detect open neural tube defects. Third, they are useful as an aid in interpreting maternal serum AFP levels when tests are performed as part of a screening program for neural tube defects.j Last, they can provide clues to the etiology of neural tube defects. Open (non-skin-covered) spina bifida is generally considered to be a more serious defect when compared to closed (skin-covered) spina bifida. Open spina bifida tends to have a higher mortality rate in the postneonatal period and greater morbidity from physical or mental handicaps.* In addition, open spina bifida is much more likely to be detected by maternal serum AFP screening programs.6 To estimate and portray visually current race- and geography-specific birth prevalence rates of open spina bifida in the United States, we developed a computergenerated mapping procedure.
Birth
prevalence
rates
of spina
bifida
571
n 8 Saredon Data from the National Center for Health Statinicr
1. Estimates of birth prevalence rates of spina bifida for white births in the United States f’rom computer-generated maps.
Fig.
MWriel
end methods
To estimate prevalence rates based on race and geography, we used birth certificate data for 1973 to 1977 from the National Center for Health Statistics (NCHS). This data set included all but those from four states (Georgia, Indiana, New Jersey, and North Carolina). Data on spina bifida were used. Based on data from Great Britain and from the Metropolitan Atlanta Congenital Defects Program (MACDP), about 75% of all cases of spina bifida represent open spina bifida.6* ’ An unpublished computer mapping procedure previously developed by L. M. James, J. D. Erickson, and P. J. Halbritter was used as the basis for these calculations. A grid of 33-mile squares superimposed on a map of the United States is used in this technique. Rates of spina bifida were computed for each square on the basis of cumulative data from counties having population centers within a loo-mile radius of the center of each square. However, squares that had no county population center within a 50-mile radius of the square were excluded from these calculations to avoid characterizing squares based only on outlying data. These data from each square were then adjusted to correct for underascertainment and decreasing secular trend in spina bifida rates. Comparison of the rate of spina bifida recorded on birth certificates with the rate of spina bifida reported in the MACDP’ and the Birth Defects Monitoring Program (BDMP)s indicates that
only about 57% of cases are reported on birth certificates. The data were thus multiplied by the factor 1.75 to adjust for underreporting. An adjustment for secular trend was made with the use of rates of spina bifida from the BDMP for each year 1973 to 1980. The data were first smoothed by fitting a regression tine to the logarithm of the rates. The ratio of the smoothed rates for 1975 (the midpoint of the birth certificate data) to that for 1980 (the target year) was 0.88, the factor by which thr NCHS rat.es were further multiplied. In addition, a factor of 0.75 was used as the ratio of open spina bifida. IO total spina bifida incidence, as noted previously. Last, in order to correct for variable rates of reporting, rates in each square were further adjusted by the ratio of the rate of all reported defects, excluding spina bifida and anencephaly, for the total NCHS data to the corresponding rate for each square. This process of computing rates for squares is anatogous to a moving average and would tend IO smooth rates. However, after this smoothing procttss, the rates still showed considerable local fluctuation. especially for areas with small numbers of births. Therefore, the rates were smoothed even further by an iterative process of averaging rates for adjacent squares. For each iteration, the “new” rate for each square was calculated as the average of the “old” rate for the set of. squares (up to nine) composed of the square itsetf‘and a+jacent
572
Greenberg, James, and Oakley
Bared
on Data
from
the
National
March 1. 1983 Am. J. Obstet. Gynecol.
Center
for
Health
Statistics
Fig. 2. Estimates of birth prevalence rates of spina bifida for black births in the United States from computer-generated maps. squares with estimated rates. This averaging process was repeated as many times as necessary in order to achieve a map with isometric features, that is, one showing broad areas with equal prevalence rates. Two maps were produced: one for white births and one for black births. The map for white births required 20 iterations to achieve final smoothing, and that for black births, 75. More iterations were necessary for black births because the sparsity of black births in certain areas produced greater fluctuations. Rates in areas with missing or sparse NCHS data were estimated with data from the BDMP and the MACDP.7, ’ The final map was then hand drawn to show the isometric prevalence regions.
Results The map of rates for white births is shown in Fig. 1. It depicts relatively high rates in the Southeast, with a maximum of eight per 10,000 total live births. This is consistent with anecdotal reports suggesting higher rates in southern Appalachia, although in some areas of Kentucky, Tennessee, and North and South Carolina, the actual rates may be even slightly higher than those calculated. The rates decrease from east to west, with the lowest rates of two to three per 10,000 in the Rocky Mountain states. The map of rates for black births is shown in Fig. 2. The rates for black births are highest along most of the Atlantic Coast and in the South Central region, with rates of three per 10,000.
Rates for black births are lowest is the Rocky Mountain states and the Pacific Northwest, with rates of less than one per 10,000. Tms, the overall variation in rates ranges from less than one per 10,000 for Northwestern black births to eight per 10,000 for southern Appalachian white births, an approximate ninefold difference.
Comment These maps show the most current available estimates of open spina bifida in the United States by race and geography. We believe that these maps are useful for several reasons. First, they provide an estimate of rates for geographic regions. Thus, state crippled children programs and local spina bifida groups can use these data for general planning purposes. Second, prevalence rates from the maps could be used to determine the feasibility of establishing a maternal serum AFP screening program for open neural tube defects in local areas. Third, the maps can be used as an aid in interpreting maternal serum AFP values. Unfortunately, these maps also have several disadvantages or limitations. First, the adjustment of NCHS rates may have produced overestimation of the rates in some areas and persistent underestimation in other areas. For example, the rate for white births in Atlanta is actually closer to nine per 10,000 than to eight per 10,000, based on data from MACDP.3 Second, the maps cannot be used to distinguish between ethnic
Birth
groups within a racial group. An example of this is the difference in rates for those of Irish ancestry versus those of Jewish background. Third, it is difficult to map rates for other racial groups, such as Hispanic and Oriental people and native Americans, because the sparcity of births in these groups in most areas results in extremely unstable estimates. Fourth, although the process of smoothing helps to define broad geographic prevalence rates, it may mask differences among smaller geographic units, such as counties. For these data, however, we suspect that local fluctuation in unsmoothed rates may mainly reflect chance variation. Last, for the purpose of AFP screening, these rates are based on prevalence at birth rather than in the second trimester. when testing would take place. No geographic or race-specific data on spontaneous abortion rates of fetuses with open spina bifida are available. Because many more anencephalic infants are born dead, it would be much more difficult to generate a similar map for anencephaly prevalence rates. Anen-
prevalence
rates
of spina
bifida
573
cephaly, like most malformations, is not routinely coded on fetal death certificates, resulting in a greater ascertainment problem. In addition, a map of anencephaly rates would have less practical and less clinical significance. Anencephaly is usually easier to diagnose with a greater degree of confidence during pregnancy by use of ultrasonography, with or without maternal serum AFP elevations, than is spina bifida.” Thus, although theoretically important, prevalenre rates of anencephaly are less clinically useful in interpreting serum AFP values. With the increasing availability of maternal serum AFP screening, the interpretation of .4FP levels in terms of the likelihood of having a fetus Affected with open spina bifida does become an important issue. The prevalence data from these maps should be quite valuable in this regard, and, until more precise data are available, the maps should provide useful estimates of open spina bifida rates by race and geography.
REFERENCES 1. Goldberg, M. F., and Oakley, G. P.: Prenatal screening for anencephaly-spina bifida: Some epidemiologic projections for a national program, in Porter, I. H., Hook E. B.,
2. 3. 4. 5.
editors: Service and Education in Medical Genetics, New York, 1979, Academic Press, Inc. pp. 55-68. Erickson, J. D.: Racial variations in the incidence of congenital malformations, Ann. Hum Genet. 39:315, 1976. Centers for Disease Control: Congenital Malformations Surveillance Report, January-December. 1979, issued December, 1980. Althouse, R., and Wald, N.: Survival and handicap of infants with spina bifida, Arch. Dis. Child. 55:845, 1980. Wald, N. J.: Workgroup Paper: The interpretation of AFP values and the effect of AFP assay performance on screening efficacy, in Gastel, B., Haddow, J. E., Fletcher, J, C.,
and Neale, A., editors: Maternal
Serum Alpha Fetopro-
tein: Issues in the Prenatal Screening and Diagnosis of Neural Tube Defects, Washington, D. C., 1980, United
States Government Printing 6. United Kingdom Collaborative
Office, pp. 29-33.
Study on alpha-fetoprotein measurement in antenatal screening for anencephaly and spina bifida in early pregnancy, Lancer 1:1323, 1977. 7. Oakley, G. P., Flynt, J. W., and Falek, A.: Community surveillance of birth defects, in Cohen, B. H., Lilienfeld, A. M. and Huang, P. C., editors: Genetic Issues in Public Health and Medicine, Springfield, Illinois, 1978. Charles C Thomas, Publisher, pp. 235-248. 8. Flynt, J. W.: Techniques for assessing teratogenic effects: Epidemiology, Environ. Health Perspect. 18: 117. 1976.
GREENBERG,
LEVY
M. JAMES,
GODFREY Atlanta,
M.D. M.S.
P. OAKLEY,
JR.,
M.D.,
M.S.P.M.
Georgia
A computer-generated mapping procedure was developed to estimate geographic and race-specific birth prevalence rates of open spina bifida. The estimates are based on birth certificate data adjusted for underascertainment. Separate maps were produced for white births and black births. For both races there is a general decreasing rate of spina biida from east to west. The highest rates are eight per 10,000 total births for whites in Southern Appalachia, and the lowest rates are less than one per 10,000 for blacks in the Rocky Mountain states and the Pacific Northwest. Until more exact data are available, these maps represent the best current available data on racial and geographic birth prevalence rates in the United States. They are useful for program planning and as an aid in interpreting maternal serum a-fetoprotein levels to detect neural tube defects. (AM. J. OBSTET.
GYNECOL.
145:570,
1983.)
THE NEURALTUBE DEFECTS anencephaly and spina bifida are among the most common and serious birth defects in the United States. The estimated birth prevalence rate of neural tube defects (the number of cases of neural tube defects at birth in a given area divided by the total number of live births and stillbirths in the area) in the early 1970s was 16 per 10,000 total births in the United States as a whole.’ However, there are known differences in rates when data are analyzed by race, geographic region, and time. The rate of neural tube defects among white births is two to three times higher than that among black births.2 Rates are higher in the eastern and southern United States than in the western United States, and there has been a decreasing secular trend in rates in both races and in all geographic regions.3 Recently, there has been an increasing need to provide more specific birth prevalence data on neural tube
From the Birth Defects Brunch, Chronic Diseases Division, Center for Environmental Health, Centers for Disease Control, Public Health S&e, United States Department of Health and Human Sentices. Received
for publication
Accepted
September
Reprint Branch, Control,
570
requests: Chronic Atlanta,
May
11, 1982.
2, 1982. G. P. Oakley, M.D., Birth Defecctr Diseases Division, Centers for Disease Georgia 30333.
defects, primarily spina bifida, with particular regard to racial and geographic distributions. Spina bifida is generally considered to be of greater concern because it is a major cause of morbidity from physical and mental handicaps among infants and children.“ Data on birth prevalence of spina bifida are useful for several reasons. First, they can be used for program planning for the care and treatment of affected infants and children. Second, they can be helpful in the decision to initiate a local maternal serum a-fetoprotein (AFP) screening program to detect open neural tube defects. Third, they are useful as an aid in interpreting maternal serum AFP levels when tests are performed as part of a screening program for neural tube defects.j Last, they can provide clues to the etiology of neural tube defects. Open (non-skin-covered) spina bifida is generally considered to be a more serious defect when compared to closed (skin-covered) spina bifida. Open spina bifida tends to have a higher mortality rate in the postneonatal period and greater morbidity from physical or mental handicaps.* In addition, open spina bifida is much more likely to be detected by maternal serum AFP screening programs.6 To estimate and portray visually current race- and geography-specific birth prevalence rates of open spina bifida in the United States, we developed a computergenerated mapping procedure.
Birth
prevalence
rates
of spina
bifida
571
n 8 Saredon Data from the National Center for Health Statinicr
1. Estimates of birth prevalence rates of spina bifida for white births in the United States f’rom computer-generated maps.
Fig.
MWriel
end methods
To estimate prevalence rates based on race and geography, we used birth certificate data for 1973 to 1977 from the National Center for Health Statistics (NCHS). This data set included all but those from four states (Georgia, Indiana, New Jersey, and North Carolina). Data on spina bifida were used. Based on data from Great Britain and from the Metropolitan Atlanta Congenital Defects Program (MACDP), about 75% of all cases of spina bifida represent open spina bifida.6* ’ An unpublished computer mapping procedure previously developed by L. M. James, J. D. Erickson, and P. J. Halbritter was used as the basis for these calculations. A grid of 33-mile squares superimposed on a map of the United States is used in this technique. Rates of spina bifida were computed for each square on the basis of cumulative data from counties having population centers within a loo-mile radius of the center of each square. However, squares that had no county population center within a 50-mile radius of the square were excluded from these calculations to avoid characterizing squares based only on outlying data. These data from each square were then adjusted to correct for underascertainment and decreasing secular trend in spina bifida rates. Comparison of the rate of spina bifida recorded on birth certificates with the rate of spina bifida reported in the MACDP’ and the Birth Defects Monitoring Program (BDMP)s indicates that
only about 57% of cases are reported on birth certificates. The data were thus multiplied by the factor 1.75 to adjust for underreporting. An adjustment for secular trend was made with the use of rates of spina bifida from the BDMP for each year 1973 to 1980. The data were first smoothed by fitting a regression tine to the logarithm of the rates. The ratio of the smoothed rates for 1975 (the midpoint of the birth certificate data) to that for 1980 (the target year) was 0.88, the factor by which thr NCHS rat.es were further multiplied. In addition, a factor of 0.75 was used as the ratio of open spina bifida. IO total spina bifida incidence, as noted previously. Last, in order to correct for variable rates of reporting, rates in each square were further adjusted by the ratio of the rate of all reported defects, excluding spina bifida and anencephaly, for the total NCHS data to the corresponding rate for each square. This process of computing rates for squares is anatogous to a moving average and would tend IO smooth rates. However, after this smoothing procttss, the rates still showed considerable local fluctuation. especially for areas with small numbers of births. Therefore, the rates were smoothed even further by an iterative process of averaging rates for adjacent squares. For each iteration, the “new” rate for each square was calculated as the average of the “old” rate for the set of. squares (up to nine) composed of the square itsetf‘and a+jacent
572
Greenberg, James, and Oakley
Bared
on Data
from
the
National
March 1. 1983 Am. J. Obstet. Gynecol.
Center
for
Health
Statistics
Fig. 2. Estimates of birth prevalence rates of spina bifida for black births in the United States from computer-generated maps. squares with estimated rates. This averaging process was repeated as many times as necessary in order to achieve a map with isometric features, that is, one showing broad areas with equal prevalence rates. Two maps were produced: one for white births and one for black births. The map for white births required 20 iterations to achieve final smoothing, and that for black births, 75. More iterations were necessary for black births because the sparsity of black births in certain areas produced greater fluctuations. Rates in areas with missing or sparse NCHS data were estimated with data from the BDMP and the MACDP.7, ’ The final map was then hand drawn to show the isometric prevalence regions.
Results The map of rates for white births is shown in Fig. 1. It depicts relatively high rates in the Southeast, with a maximum of eight per 10,000 total live births. This is consistent with anecdotal reports suggesting higher rates in southern Appalachia, although in some areas of Kentucky, Tennessee, and North and South Carolina, the actual rates may be even slightly higher than those calculated. The rates decrease from east to west, with the lowest rates of two to three per 10,000 in the Rocky Mountain states. The map of rates for black births is shown in Fig. 2. The rates for black births are highest along most of the Atlantic Coast and in the South Central region, with rates of three per 10,000.
Rates for black births are lowest is the Rocky Mountain states and the Pacific Northwest, with rates of less than one per 10,000. Tms, the overall variation in rates ranges from less than one per 10,000 for Northwestern black births to eight per 10,000 for southern Appalachian white births, an approximate ninefold difference.
Comment These maps show the most current available estimates of open spina bifida in the United States by race and geography. We believe that these maps are useful for several reasons. First, they provide an estimate of rates for geographic regions. Thus, state crippled children programs and local spina bifida groups can use these data for general planning purposes. Second, prevalence rates from the maps could be used to determine the feasibility of establishing a maternal serum AFP screening program for open neural tube defects in local areas. Third, the maps can be used as an aid in interpreting maternal serum AFP values. Unfortunately, these maps also have several disadvantages or limitations. First, the adjustment of NCHS rates may have produced overestimation of the rates in some areas and persistent underestimation in other areas. For example, the rate for white births in Atlanta is actually closer to nine per 10,000 than to eight per 10,000, based on data from MACDP.3 Second, the maps cannot be used to distinguish between ethnic
Birth
groups within a racial group. An example of this is the difference in rates for those of Irish ancestry versus those of Jewish background. Third, it is difficult to map rates for other racial groups, such as Hispanic and Oriental people and native Americans, because the sparcity of births in these groups in most areas results in extremely unstable estimates. Fourth, although the process of smoothing helps to define broad geographic prevalence rates, it may mask differences among smaller geographic units, such as counties. For these data, however, we suspect that local fluctuation in unsmoothed rates may mainly reflect chance variation. Last, for the purpose of AFP screening, these rates are based on prevalence at birth rather than in the second trimester. when testing would take place. No geographic or race-specific data on spontaneous abortion rates of fetuses with open spina bifida are available. Because many more anencephalic infants are born dead, it would be much more difficult to generate a similar map for anencephaly prevalence rates. Anen-
prevalence
rates
of spina
bifida
573
cephaly, like most malformations, is not routinely coded on fetal death certificates, resulting in a greater ascertainment problem. In addition, a map of anencephaly rates would have less practical and less clinical significance. Anencephaly is usually easier to diagnose with a greater degree of confidence during pregnancy by use of ultrasonography, with or without maternal serum AFP elevations, than is spina bifida.” Thus, although theoretically important, prevalenre rates of anencephaly are less clinically useful in interpreting serum AFP values. With the increasing availability of maternal serum AFP screening, the interpretation of .4FP levels in terms of the likelihood of having a fetus Affected with open spina bifida does become an important issue. The prevalence data from these maps should be quite valuable in this regard, and, until more precise data are available, the maps should provide useful estimates of open spina bifida rates by race and geography.
REFERENCES 1. Goldberg, M. F., and Oakley, G. P.: Prenatal screening for anencephaly-spina bifida: Some epidemiologic projections for a national program, in Porter, I. H., Hook E. B.,
2. 3. 4. 5.
editors: Service and Education in Medical Genetics, New York, 1979, Academic Press, Inc. pp. 55-68. Erickson, J. D.: Racial variations in the incidence of congenital malformations, Ann. Hum Genet. 39:315, 1976. Centers for Disease Control: Congenital Malformations Surveillance Report, January-December. 1979, issued December, 1980. Althouse, R., and Wald, N.: Survival and handicap of infants with spina bifida, Arch. Dis. Child. 55:845, 1980. Wald, N. J.: Workgroup Paper: The interpretation of AFP values and the effect of AFP assay performance on screening efficacy, in Gastel, B., Haddow, J. E., Fletcher, J, C.,
and Neale, A., editors: Maternal
Serum Alpha Fetopro-
tein: Issues in the Prenatal Screening and Diagnosis of Neural Tube Defects, Washington, D. C., 1980, United
States Government Printing 6. United Kingdom Collaborative
Office, pp. 29-33.
Study on alpha-fetoprotein measurement in antenatal screening for anencephaly and spina bifida in early pregnancy, Lancer 1:1323, 1977. 7. Oakley, G. P., Flynt, J. W., and Falek, A.: Community surveillance of birth defects, in Cohen, B. H., Lilienfeld, A. M. and Huang, P. C., editors: Genetic Issues in Public Health and Medicine, Springfield, Illinois, 1978. Charles C Thomas, Publisher, pp. 235-248. 8. Flynt, J. W.: Techniques for assessing teratogenic effects: Epidemiology, Environ. Health Perspect. 18: 117. 1976.