January 1979
26
The Journal o f P E D I A T R I C S
Percentile curves for hemoglobin and red cell volume in infancy and childhood Percentile curves were calculated for hemoglobin and mean corpuscular volume in children between 0.5 and 16 years of age. The curves were derived from several populations of non-indigent white children who lived near sea level Subjects were excluded from the reference population if they had laboratory evidence of iron deficiency, thalassemia minor, and/or hemoglobinopathy. The final reference populations included 9,946 children for the derivation of the hemoglobin curves and 2,314for the M C V curves. The percentile curves should be particularly applicable to the diagnosis and screening of iron deficiency and thalassemia minor.
Peter R. D a l l m a n , M.D.,* S a n Francisco, Calif., and M a r t t i A. Siimes, M . D . , H e l s i n k L F i n l a n d
I R O N D E F I C I E N C Y is by far the most common cause of
a subnormal hemoglobin concentration; among many ethnic groups, thalassemia minor is next in incidence. In each condition, anemia is usually mild and there is substantial overlap into the normal range. Iron deficiency and thalassemia minor are also characterized by a decrease in red cell volume. With the recent introduction of electronic counters into large clinical laboratories, it has become practical to determine the concentration of hemoglobin and the red cell indices concurrently. By adding the evaluation of mean corpuscular volume to the hemoglobin determination, the reliability of diagnosis can be increased substantially. Values for both hemoglobin and MCV undergo marked changes during development. In order to obtain reliable reference standards, the reference population must be screened to exclude iron deficiency and such relatively common conditions as thalassemia minor. It is also necessary to take into consideration the consistently lower mean concentration of hemoglobin in blacks than in whites and Orientals (about 0.5 gm/dl) that appears to be From the Department of Pediatrics, University of California, san Francisco, and the Children's Hospital, University of Helsinki. Supported by grants from the National Institutes of Health, Grant No. A M HD 13897, and the Foundation for Pediatric Research in Finland. *Reprints address: 650-M, Universityof California Medical Center, San Francisco, CA 94143.
Vol. 94, No. l, pp. 26-31
independent of iron deficiency and thalassemia minor.' Reference values for concentration of hemoglobin will be lowered in proportion to the percentage of blacks that are included in the reference population.
Abbreviation used MCV: mean corpuscular volume
Most tabulations of hemoglobin concentration during childhood show increases of 0.5 to 1.0 gm/dl from one age range to the next. In screening healthy infants and children, an error of 0.5 gm/dl can result in 10% of normal individuals being incorrectly categorized as anemic. Frerichs et al'-' recently calculated the cost of ignoring the relatively small hemoglobin difference between blacks and whites. In a small Louisiana community, the cost of following up the results of screening for iron deficiency in 1,000 white children would be $3,320, whereas $10,000 would be spent for 1,000 black children using the same criterion for normal hemoglobin; the actual incidence of iron deficiency was estimated to be equal in the two groups. The incorrect classifying of "anemic" individuals as normal is equally important. In order to make optimal use of recent normative data, we developed percentile grids for hemoglobin and m e a n corpuscular volume versus age which are similar to the familiar grids for height and weight. The curves were derived from several populations of white children who
0022-3476/79/100026+06500.60/0 9 1979 The C. V. Mosby Co.
Volume 94
Percentile curves for hemoglobin and red cell volume
27
Number I
lived near sea level and who did not include indigent subjects. In all but one of the groups, additional laboratory data allowed the exclusion of subjects with presumptive evidence of iron deficiency, thalassemia minor, and/or hemoglobinopathy.
17
16 15
METHODS All values for hemoglobin and MCV were obtained on venous blood by electronic counter (Coulter model S). Hemoglobin values were derived from a total of 9,946 children, and MCV values from 2,314 children (after excluding individuals who did not meet the criteria to be included in the reference population). In the case of hemoglobin, values for girls and boys were combined between the ages of 0.5 and 9 years after finding no consistent difference in concentration between the two sexes. Data for MCV for boys and girls below the age of 7 were combined on the same basis. The characteristics of the various groups were as follows. United States. US I. This group of 1,358 white children, belonging to a prepaid health plan, had a multiphasic health examination at the Kaiser Permanente Out-Patient Clinic in San Francisco between 1973 and 1975. The ages were calculated to the nearest birthday and ranged from 5 to 14 years. For the calculation of hemoglobin percentiles, all subjects with mean corpuscular volumes more than 5% below the mezn for the same age and sex were excluded in order to eliminate subjects most likely to have iron deficiency or thalassemia minor. This criterion excluded 9.8% of the subjects. For the calculation of MCV percentiles, all subjects with hemoglobin concentrations below the twentieth percentile for the same age and sex were excluded to eliminate subjects most likely to have iron deficiency or thalassemia minor. Hemoglobin electrophoresis was also used to exclude a very small number of subjects (0.2% of the total) with hemoglobin S or C trait.' US IL This group of 7,489 white children was evaluated in the Kaiser Permanente Multiphasic Program in San Francisco, between the years 1970 and 1973. Ages for this group were calculated to the nearest year and ranged from 5 through 16 years. Since the data for mean corpuscular volume and hemoglobin electrophoresis could not be readily retrieved, percentiles for hemoglobin were calculated from the entire group. US III. This group is comprised of 210 white children seen at the Moffitt Hospital Out-Patient Clinic in San Francisco between 1974 and 1977. They fell into three age categories: 10 to 17 months of age, 11/2to 4 years, and 4 to 7 years. For calculations of hemoglobin values, all those
14 13 G/D[
2
4
6
8
10
12
14
16
AGE, YEARS
Fig. 1. Hemoglobin concentration in boys. The third, fiftieth, and ninty-seventh percentile curves and the individual points from which they were derived are shown,
with serum ferritin values below 10 ng/ml, transferrin saturation below 16%, and low MCV for age (less than 70 fl, less than 73 fl, and less than 75 fl in the three age groups, respectively):~ were excluded. These criteria resulted in 27% of the subjects being excluded, primarily from the youngest age group and most frequently on the basis of a low serum ferritin value, leaving a total of 158 children. MCV values were calculated after excluding those subjects whose hemoglobin values were more than 2 SD below the mean by the above criteria, or who had serum ferritin values below 10 ng/ml, transferrin saturation below 16%, or abnormal hemoglobin electrophoresis. Most of the subjects were included in one or both earlier reports on developmental changes in the MCV ~ and serum iron.; Although 16% is commonly used as the lower limit of normal for transferrin saturation in adults, the corresponding value was recently reported to be 10% in infants" and 7% in children below 12 years of age:'; in both studies, subjects with low hemoglobin, MCV, and serum ferritin values had been excluded. Nevertheless, we used the value of 16% because of the poor reproducibility of the serum iron concentration and the likelihood that some iron-deficient individuals would have values overlapping into the normal range. Finland. F 1. This group includes 777 children ~, about half of whom were from a middle-class suburb in the greater Helsinki area (Espoo); the other half were from a smaller community in an agricultural area of central Finland (Kiuruvesi). The suburban children constituted 50% and
28
Dallman and Siimes
The Journal q[ Pediatrics January 1979
GIRLS
G/DL 12
11
8~ 8C FL 7~ 7(;
|
2
4
6
8 10 AGE, YEARS
12
14
16
(~)
2
4
6
8 10 AGE, YEARS
12
14
16
Fig. 2. A, Hemoglobin and MCV percentile curves for girls. B, Hemoglobin and MCV percentile curves for boys. the rural children 90% of the total populations of that age in their respective communities. Children were studied in 1975 at age 2, 4, 7, 10, and 15 years to the nearest birthday. For the calculation of hemoglobin values, 168 subjects were excluded because they had a serum ferritin value less than 10 ng/ml, a transferrin saturation less than 16%, or a low MCV for age (less than 73 fl at age 2, less than 75 at age 4, less than 76 at age 10, and less than 78 at age 15 years)? For the calculation of MCV values, subjects were excluded if the value for hemoglobin was more than 2 SD below the normal for the same age and sex after the above exclusions, if the serum ferritin was less than l0 ng/ml, or if the transferrin saturation was less than 16%. F 11. In this group, infants were evaluated longitudinally, with 238, 228, and 238 subjects sampled at 6, 9, and 12 months (to the nearest month), respectively." ~ They were seen at the well-baby clinic of the Helsinki Children's Hospital. The criteria for excluding subjects prior to the calculations of hemoglobin and MCV were the same as in F I, except that subjects with an MCV below 70 were excluded for the calculation of hemoglobin values. These criteria resulted in the exclusion of 83, 69, and 84 infants at 6, 9, and 12 months, respectively. These exclusions were primarily on the basis of transferrin saturation.
STATISTICAL
METHODS
All data were converted to percentiles with the e~ception of US IlI, in which the groups were small and the mean _~ 2 SD were considered equivalent to the fiftieth, ninety-seventh, and third percentiles. Potentially iron-deficient or thalassemic individuals could not be reliably excluded from group US II. Nevertheless, these data were included because US If had the advantage of being a very large group (7,489) of older children who were beyond the peak age for iron defi~ency, The fiftieth percentile values were very similar to or actually higher than those of an analogous population (US l) that was studied under identical circumstances in 1973 to 1976 instead of 1970 to 1973. In analyzing the data from US I, we found that exclusion of subjects who did not meet the criteria had virtually no effect on the median hemoglobin value. Thus inclusion of US I1 was not felt to be likely to exert a detectable downward bias in the data, except perhaps at the third percentiles. RESULTS The derivation of third, fiftieth, and ninety-seventh percentile curves for hemoglobin in boys is shown in Fig. l, in which different symbols are used for the results from
Volume 94 Number I
the various groups. For clarity of presentation, the data for the tenth and ninetieth percentiles are not shown. There is very good general agreement in the fiftieth percentile values from the various populations. Hemoglobin values in girls and MCV in both sexes were similarly in close accord (complete figures corresponding to Fig. 1 are not included here but will be supplied on request). Since one or two aberrant values can exert a strong influence on the third and ninety-seventh percentile values, it is not surprising to find more scatter among these points. Percentile curves are shown without individual data points in Fig. 2. The hemoglobin and MCV are shown together for each sex to facilitate interpretation of both laboratory results for individual patients. Values for medians and lower limits of normal by age range were derived from the fiftieth and third percentile curves, respectively, and tabulated in the Table as an alternative to the developmental grids. Values for MCV are plotted in a similar manner. Again, there is close agreement among groups, especially at the fiftieth percentile. Values for boys and girls are virtually identical, except that the rise in MCV occurs at a somewhat earlier age in girls than in boys. There is greater variability at the third and ninety-seventh percentiles. Values for MCV rise very gradually through most of childhood, with a median value of 84 to 85 fl at 15 years of age that is still well below the adult value of about 90? Some data indicate that MCV values do not reach a plateau even in the adult, but that they continue to increase throughout life. ~~ 11 DISCUSSION There are two general approaches to establishing criteria of normality; one involves studying a random sample of the total population. This approach is exemplified by the recently published growth grids from the National Center for Health Statistics that provide a standard method for assessing growth of contemporary infants, children, and youth in the United States. 1..... One disadvantage &these grids is that they do not exclude values of abnormal individuals. For example, body weight curves reflect current conditions, with obesity a common abnormality. Inclusion of obese individuals in the derivation of weight grids may be misleading if the curves are interpreted as representing optimal standards. However, exclusion of obese individuals or other abnormal categories involves applying somewhat arbitrary definitions. Consequently, the use of height and weight values from a random sample of the population is the least controversial method of obtaining growth standards. A second group of methods for deriving normative data involves the exclusion of all subjects with evidence of
Percentile curves for hemoglobin and red cell volume
29
Table
Hemoglobin (gm/ dl) Age @r) 0.5-2 2-5 5-9 9-12 12-14 Female Male 14-18 Female Male
MCV (fl)
Median
Lower limit
Median
Lower limit
12.5 12.5 13.0 13.5
11,0 11.0 11.5 12.0
77 79 81 83
70 73 75 76
13.5 14.0
12.0 12.5
85 84
77 76
14.0 15.0
12.0 13.0
87 86
78 77
14.0 16.0
12.0 14.0
90 90
80 80
18-49
Female Male
common abnormalities or adverse environmental circumstances that might be expected to bias the data. Although this method involves the use of criteria that might seem arbitrary, it has certain advantages for the derivation of normative data for hemoglobin and mean corpuscular volume. Iron deficiency is particularly prevalent between the ages of I and 3, '~ and almost any sampling of this age group will include substantial numbers of children with iron deficiency anemia. The prevalence of thalassemia minor will vary according to the ethnic make-up of the population. Since normal standards are to be applied in screening for these two common conditions, it is desirable that the standards are calculated after excluding as many of these individuals as possible. The curves in the Figures represent an initial attempt to do this. Any criteria that reliably exclude iron deficiency also exclude some iron-sufficient subjects. Indeed, the more sensitive the criteria are in respect to excluding iron deficiency, the less specific they are likely to be by also excluding a larger percentage of the iron-sufficient population. Since it is likely that substantial numbers of iron-sufficient children were excluded by our criteria, it is important to ask whether the removal of this group biased the reference data. In the derivation of normal hemoglobin values, for example, it was assumed that hemoglobin concentration and MCV are dependent variables only when a condition such as iron deficiency or thalassemia minor results in both laboratory values being low. In the reference population, hemoglobin and MCV were considered to be independent variables. Although the latter is not strictly correct, it is close enough to the truth to be a reasonable working assumption. This was evident in US I, in which there was a very slight but significant tendency
30
Dallman and Siimes
for high MCV values to be associated with high hemoglobin concentration at any given age.' Based on normalized data from the entire group, we would predict that 10year-old white children, for example, with a higher than average MCV of 86 fl would have a mean hemoglobin of 13.5 gm/dl; with a lower than average MCV of 78 fl there would be a slightly lower hemoglobin of 13.4 gm/dl. Only when the MCV fell below the presumptive lower limit of normal of 76 (roughly corresponding to our exclusion criterion) was there a much steeper and more substantial fall in mean hemoglobin to 12.5 gm/dl. We felt it legitimate to assume that MCV and hemoglobin were independent variables within the reference population because the hemoglobin concentrations were unlikely to be altered by more than 0.1 gm/dl by excluding normal subjects on the basis of our MCV criteria. The same reasoning applies to the calculation of MCV after exclusion of low hemoglobin values. It could also be asked whether our criteria for transferfin saturation and serum ferritin would bias the calculation of reference values for hemoglobin and MCV. In iron deficiency, low transferrin saturation and serum ferritin concentrations are associated with low values of hemoglobin and MCV, but there is no evidence for a similar correlation in iron-sufficient individuals. We compared the mean serum iron in screened children of the three age groups in US III whose hemoglobin, MCV, and serum ferritin values, respectively, were below the median for each test with those whose values were above the median.' We found no significant difference in serum iron between the high and low hemoglobin, MCV, and serum ferritin values. US I was previously part of a study of hemoglobin values in whites, blacks, and Asians. As had been found in other surveys, 1:'-~ blacks were found to have consistently lower hemoglobin values by about 0.5 gm/dl.' This difference should probably be taken into account in using these grids. Data from a smaller number of Asians of various ethnic backgrounds (primarily Chinese, Japanese, and Filipino) were also analyzed, and there was no evidence that their values were different from those of whites after subjects with low MCV values were excluded. MCV values in blacks may also be somewhat lower than in whites? but the magnitude of this difference after exclusion of thalassemia minor and iron deficiency has not been determined. A comparison with earlier published values indicates that hemoglobin concentrations in this paper are higher than previous tabulations of normal values at many ages. Values at one year of age would be expected to be most subject to error if they are based on a population with a high prevalence of subclinical iron deficiency. In the large
The Journal of Pediatrics January 1979
study of Guest and Brown, '~ the mean hemoglobin concentration at 1 year of age was 11.1 gm/dl. Subsequent tabulations from textbooks and surveys range from 11.2 to 12.0 gm/dl}' .~0-24all substantially lower than our median value of 12.5 g m / d l ? Hunter and Smith z~ demonstrated that exclusion of subjects with laboratory evidence of iron deficiency substantially raised the distribution of hemoglobin and hematocrit values in a group of healthy infants of various racial groups. The fact that our values are derived from a white population is an additional explanation for the finding of higher values than in studies based on a more racially heterogeneous population. Alternative methods of deriving normative data for hemoglobin involve excluding iron-deficient subjects by employing a therapeutic trial of iron S. '-'~or assuming that hemoglobin values in healthy populations follow a single Gaussian distribution/'~ During adolescence, developmental changes in hemoglobin or hematocrit are a function of the stage of sexual maturation in boys. ~ It is possible to develop criteria of normal for each stage of puberty, but the use of percentile grids by age seems more practical for routine clinical use. Differences in onset of puberty account for the large spread between the third and ninety-seventh percentiles for hemoglobin in boys. For most laboratory studies, values are considered subnormal when they are more than 2 SD below the mean~ a lower limit that corresponds to the third percentile. In the case of hemoglobin, it is recognized that there is a large overlap between normal and subnormal values. ~..... For example, an individual may have a value that lies in the lower portion of the normal distribution for his sex and age despite having mild iron deficiency. This individual, whose hemoglobin or hematocrit would normally be higher, would not be recognized as being "anemic" unless he were treated with iron. '-'~By combining the MCV with the hemoglobin in screening for iron deficiency, '~ the accuracy of diagnosis is increased because it is less likely that both values would be in the lowest portion of the distribution curve by chance.' In populations in which iron deficiency is common, values for hemoglobin and MCV below the tenth percentile might be considered grounds for suspecting iron deficiency. However, in a population that contains few iron-deficient individuals, criteria should probably be more stringent because it is difficult to justify investigating too many normal children in the lower part of the normal distribution curve in order to discover every example of mild deficiency. The statistical considerations in evaluating these two types of populations are applicable to many screening situations? ~ The beneficial results of treating iron deficiency and counseling individuals with
Volume 94 Number 1
thalassemia m i n o r m u s t be b a l a n c e d against the cost a n d inconvenience of the screening p r o c e d u r e to n o r m a l subjects. Our d e v e l o p m e n t a l grids for h e m o g l o b i n a n d M C V are a n attempt to deal with currently available data. As more c o m p r e h e n s i v e survey data b e c o m e available, a n d with
Percentile curves for hemoglobin and red cell volume
14. 15.
16.
verification o f reference values by t h e r a p e u t i c trial with iron, it is likely that a d j u s t m e n t s a n d modifications will
17.
have to be made. REFERENCES
1. Dallman PR, Barr GD, Allen CM, and Shinefield HR: Hemoglobin concentration in white, black, and Oriental children: is there a need for separate criteria in screening for anemia? Am J Clin Nutr 31:377, 1978. 2. Frerichs RR, Webber LS, Srinivasan SR, and Berenson GS: Hemoglobin levels in children from a biracial southern community, Am J Public Health 67:841, 1977. 3. Dallman PR: Nutritional anemias, in Handbook of clinical nutrition, Evanston, American Academy of Pediatrics (in press). 4. Koerper MA, Mentzer WC, Brecher (3, and Dallman PR: Developmental change in red blood cell volume: Implications in screening infants and children for iron deficiency and thalassemia trait, J PEDIATR 89:580, 1976. 5. Koerper MA, and Dallman PR: Serum iron concentration and transferrin saturation in the diagnosis of iron deficiency in children: Normal developmental changes, J P~DIATR 91:870, 1977. 6. Saarinen UM, and Siimes MA: Developmental changes in serum iron, total iron-binding capacity and transferrin saturation in infancy, J PEDIATR 91:875, 1977. 7. Slimes MA, and Lundstr6m U: Unpublished data. 8. Saarinen UM, and Siimes MA: Developmental changes in the red blood cell counts and indices in infants with iron deficiency excluded both by laboratory criteria and by continuous iron supplementation, J PEDIATR 92:412, 1978. 9. Williams WJ, Beutler E, Erslev AL, and Rundles RW: Hematology, ed 2, 1977, McGraw-Hill Book Company, New York, p 10. 10. Okuno T: Red cell size as measured by the Coulter model S, J Clin Pathol 25:599, 1972. I1. Okuno T: Red cell size and age. Br Med J 1:569, 1972. 12. National Center for Health Statistics: NCHS growth charts, 1976, Monthly Vital Statistics Report, Vol 25, No 3, Suppl (HRA) 76-1120. 13. National Center for Health Statistics: NCHS growth charts, 1976, Vital and Health Statistics, Series 11, Health Resources Administration, Rockville, Md., U.S. Government Printing Office.
18.
19.
20. 21.
22.
23. 24.
25.
26. 27. 28.
29. 30. 31.
31
Lahey ME: Iron deficiency anemia, Pediatr Clin North Am 4:481, 1957. Ten-State Nutrition Survey, U.S. Department of Health, Education and Welfare, DHEW Publ. No. (HSM) 72-8130, 1972. Owen GM, Lubin AH, and Garry PJ: Hemoglobin levels according to age, race, and transferrin saturation in preschool children of comparable socioeconomic status, J PEDIATR 82:850, 1973. Garn SM, Smith NJ, and Clark DC: Lifelong differences in hemoglobin levels between blacks and whites, J Natl Med Assoc 67:91, 1975. Owen GM, and Yanochik-Owen A: Should there be a different definition of anemia in black and white children? Am J Public Health 67:865, 1977. Guest GM, and Brown EW: Erythrocytes and hemoglobin of the blood in infancy and childhood, III. Factors in variability, statistical studies, Am J Dis Child 93:486, 1957. Moe PJ: Normal red blood picture during the first 3 years of life, Acta Pediatr Scand 54:69, 1965. Burman D: Haemoglobin levels in normal infants aged 3 to 24 months and the effect of iron, Arch Dis Child 47:261, 1972. Hunter RE, and Smith N J: Hemoglobin and hematocrit values in iron deficiency in infancy, J PEDIATR 81:710, 1972. Wintrobe MM: Clinical hematology, ed 7, Philadelphia, 1974, Lea & Febiger, Publishers, p 1797. Dallman PR: Blood and blood forming tissues, in Rudolph AM, editor: Pediatrics, ed 16, Appleton-Century-Crofts, Inc, New York, 1977, p 1111. Garby L, Irnell L, and Werner I: Iron deficiency in women of fertile age in a Swedish community. II. Efficiency of several laboratory tests to predict the response to iron supplementation, Acta Med Scand 185:107, 1969. Cook JD, et al: Nutritional deficiency and anemia in Latin America, Blood 38:591, 1971. Daniel WA: Hematocrit: maturity relationship in adolescence, Pediatrics 52:388, 1973. Garby L, Irnell L, and Werner I: Iron deficiency in a Swedish community. III. Estimation of prevalence based on response to iron supplementation, Acta Med Scand 185:113, 1969. Cook JD, Finch CA, and Smith N J: Evaluation of the iron status of a population, Blood 48:449, 1976. Dallman PR: New approaches to screening for iron deficiency. J PEDIATR 90:678, 1977. Galen RS, and Gambino SR: Beyond normality: The predictive value and efficiency of medical diagnoses, New York, 1975, John Wiley & Sons, Inc.
26
The Journal o f P E D I A T R I C S
Percentile curves for hemoglobin and red cell volume in infancy and childhood Percentile curves were calculated for hemoglobin and mean corpuscular volume in children between 0.5 and 16 years of age. The curves were derived from several populations of non-indigent white children who lived near sea level Subjects were excluded from the reference population if they had laboratory evidence of iron deficiency, thalassemia minor, and/or hemoglobinopathy. The final reference populations included 9,946 children for the derivation of the hemoglobin curves and 2,314for the M C V curves. The percentile curves should be particularly applicable to the diagnosis and screening of iron deficiency and thalassemia minor.
Peter R. D a l l m a n , M.D.,* S a n Francisco, Calif., and M a r t t i A. Siimes, M . D . , H e l s i n k L F i n l a n d
I R O N D E F I C I E N C Y is by far the most common cause of
a subnormal hemoglobin concentration; among many ethnic groups, thalassemia minor is next in incidence. In each condition, anemia is usually mild and there is substantial overlap into the normal range. Iron deficiency and thalassemia minor are also characterized by a decrease in red cell volume. With the recent introduction of electronic counters into large clinical laboratories, it has become practical to determine the concentration of hemoglobin and the red cell indices concurrently. By adding the evaluation of mean corpuscular volume to the hemoglobin determination, the reliability of diagnosis can be increased substantially. Values for both hemoglobin and MCV undergo marked changes during development. In order to obtain reliable reference standards, the reference population must be screened to exclude iron deficiency and such relatively common conditions as thalassemia minor. It is also necessary to take into consideration the consistently lower mean concentration of hemoglobin in blacks than in whites and Orientals (about 0.5 gm/dl) that appears to be From the Department of Pediatrics, University of California, san Francisco, and the Children's Hospital, University of Helsinki. Supported by grants from the National Institutes of Health, Grant No. A M HD 13897, and the Foundation for Pediatric Research in Finland. *Reprints address: 650-M, Universityof California Medical Center, San Francisco, CA 94143.
Vol. 94, No. l, pp. 26-31
independent of iron deficiency and thalassemia minor.' Reference values for concentration of hemoglobin will be lowered in proportion to the percentage of blacks that are included in the reference population.
Abbreviation used MCV: mean corpuscular volume
Most tabulations of hemoglobin concentration during childhood show increases of 0.5 to 1.0 gm/dl from one age range to the next. In screening healthy infants and children, an error of 0.5 gm/dl can result in 10% of normal individuals being incorrectly categorized as anemic. Frerichs et al'-' recently calculated the cost of ignoring the relatively small hemoglobin difference between blacks and whites. In a small Louisiana community, the cost of following up the results of screening for iron deficiency in 1,000 white children would be $3,320, whereas $10,000 would be spent for 1,000 black children using the same criterion for normal hemoglobin; the actual incidence of iron deficiency was estimated to be equal in the two groups. The incorrect classifying of "anemic" individuals as normal is equally important. In order to make optimal use of recent normative data, we developed percentile grids for hemoglobin and m e a n corpuscular volume versus age which are similar to the familiar grids for height and weight. The curves were derived from several populations of white children who
0022-3476/79/100026+06500.60/0 9 1979 The C. V. Mosby Co.
Volume 94
Percentile curves for hemoglobin and red cell volume
27
Number I
lived near sea level and who did not include indigent subjects. In all but one of the groups, additional laboratory data allowed the exclusion of subjects with presumptive evidence of iron deficiency, thalassemia minor, and/or hemoglobinopathy.
17
16 15
METHODS All values for hemoglobin and MCV were obtained on venous blood by electronic counter (Coulter model S). Hemoglobin values were derived from a total of 9,946 children, and MCV values from 2,314 children (after excluding individuals who did not meet the criteria to be included in the reference population). In the case of hemoglobin, values for girls and boys were combined between the ages of 0.5 and 9 years after finding no consistent difference in concentration between the two sexes. Data for MCV for boys and girls below the age of 7 were combined on the same basis. The characteristics of the various groups were as follows. United States. US I. This group of 1,358 white children, belonging to a prepaid health plan, had a multiphasic health examination at the Kaiser Permanente Out-Patient Clinic in San Francisco between 1973 and 1975. The ages were calculated to the nearest birthday and ranged from 5 to 14 years. For the calculation of hemoglobin percentiles, all subjects with mean corpuscular volumes more than 5% below the mezn for the same age and sex were excluded in order to eliminate subjects most likely to have iron deficiency or thalassemia minor. This criterion excluded 9.8% of the subjects. For the calculation of MCV percentiles, all subjects with hemoglobin concentrations below the twentieth percentile for the same age and sex were excluded to eliminate subjects most likely to have iron deficiency or thalassemia minor. Hemoglobin electrophoresis was also used to exclude a very small number of subjects (0.2% of the total) with hemoglobin S or C trait.' US IL This group of 7,489 white children was evaluated in the Kaiser Permanente Multiphasic Program in San Francisco, between the years 1970 and 1973. Ages for this group were calculated to the nearest year and ranged from 5 through 16 years. Since the data for mean corpuscular volume and hemoglobin electrophoresis could not be readily retrieved, percentiles for hemoglobin were calculated from the entire group. US III. This group is comprised of 210 white children seen at the Moffitt Hospital Out-Patient Clinic in San Francisco between 1974 and 1977. They fell into three age categories: 10 to 17 months of age, 11/2to 4 years, and 4 to 7 years. For calculations of hemoglobin values, all those
14 13 G/D[
2
4
6
8
10
12
14
16
AGE, YEARS
Fig. 1. Hemoglobin concentration in boys. The third, fiftieth, and ninty-seventh percentile curves and the individual points from which they were derived are shown,
with serum ferritin values below 10 ng/ml, transferrin saturation below 16%, and low MCV for age (less than 70 fl, less than 73 fl, and less than 75 fl in the three age groups, respectively):~ were excluded. These criteria resulted in 27% of the subjects being excluded, primarily from the youngest age group and most frequently on the basis of a low serum ferritin value, leaving a total of 158 children. MCV values were calculated after excluding those subjects whose hemoglobin values were more than 2 SD below the mean by the above criteria, or who had serum ferritin values below 10 ng/ml, transferrin saturation below 16%, or abnormal hemoglobin electrophoresis. Most of the subjects were included in one or both earlier reports on developmental changes in the MCV ~ and serum iron.; Although 16% is commonly used as the lower limit of normal for transferrin saturation in adults, the corresponding value was recently reported to be 10% in infants" and 7% in children below 12 years of age:'; in both studies, subjects with low hemoglobin, MCV, and serum ferritin values had been excluded. Nevertheless, we used the value of 16% because of the poor reproducibility of the serum iron concentration and the likelihood that some iron-deficient individuals would have values overlapping into the normal range. Finland. F 1. This group includes 777 children ~, about half of whom were from a middle-class suburb in the greater Helsinki area (Espoo); the other half were from a smaller community in an agricultural area of central Finland (Kiuruvesi). The suburban children constituted 50% and
28
Dallman and Siimes
The Journal q[ Pediatrics January 1979
GIRLS
G/DL 12
11
8~ 8C FL 7~ 7(;
|
2
4
6
8 10 AGE, YEARS
12
14
16
(~)
2
4
6
8 10 AGE, YEARS
12
14
16
Fig. 2. A, Hemoglobin and MCV percentile curves for girls. B, Hemoglobin and MCV percentile curves for boys. the rural children 90% of the total populations of that age in their respective communities. Children were studied in 1975 at age 2, 4, 7, 10, and 15 years to the nearest birthday. For the calculation of hemoglobin values, 168 subjects were excluded because they had a serum ferritin value less than 10 ng/ml, a transferrin saturation less than 16%, or a low MCV for age (less than 73 fl at age 2, less than 75 at age 4, less than 76 at age 10, and less than 78 at age 15 years)? For the calculation of MCV values, subjects were excluded if the value for hemoglobin was more than 2 SD below the normal for the same age and sex after the above exclusions, if the serum ferritin was less than l0 ng/ml, or if the transferrin saturation was less than 16%. F 11. In this group, infants were evaluated longitudinally, with 238, 228, and 238 subjects sampled at 6, 9, and 12 months (to the nearest month), respectively." ~ They were seen at the well-baby clinic of the Helsinki Children's Hospital. The criteria for excluding subjects prior to the calculations of hemoglobin and MCV were the same as in F I, except that subjects with an MCV below 70 were excluded for the calculation of hemoglobin values. These criteria resulted in the exclusion of 83, 69, and 84 infants at 6, 9, and 12 months, respectively. These exclusions were primarily on the basis of transferrin saturation.
STATISTICAL
METHODS
All data were converted to percentiles with the e~ception of US IlI, in which the groups were small and the mean _~ 2 SD were considered equivalent to the fiftieth, ninety-seventh, and third percentiles. Potentially iron-deficient or thalassemic individuals could not be reliably excluded from group US II. Nevertheless, these data were included because US If had the advantage of being a very large group (7,489) of older children who were beyond the peak age for iron defi~ency, The fiftieth percentile values were very similar to or actually higher than those of an analogous population (US l) that was studied under identical circumstances in 1973 to 1976 instead of 1970 to 1973. In analyzing the data from US I, we found that exclusion of subjects who did not meet the criteria had virtually no effect on the median hemoglobin value. Thus inclusion of US I1 was not felt to be likely to exert a detectable downward bias in the data, except perhaps at the third percentiles. RESULTS The derivation of third, fiftieth, and ninety-seventh percentile curves for hemoglobin in boys is shown in Fig. l, in which different symbols are used for the results from
Volume 94 Number I
the various groups. For clarity of presentation, the data for the tenth and ninetieth percentiles are not shown. There is very good general agreement in the fiftieth percentile values from the various populations. Hemoglobin values in girls and MCV in both sexes were similarly in close accord (complete figures corresponding to Fig. 1 are not included here but will be supplied on request). Since one or two aberrant values can exert a strong influence on the third and ninety-seventh percentile values, it is not surprising to find more scatter among these points. Percentile curves are shown without individual data points in Fig. 2. The hemoglobin and MCV are shown together for each sex to facilitate interpretation of both laboratory results for individual patients. Values for medians and lower limits of normal by age range were derived from the fiftieth and third percentile curves, respectively, and tabulated in the Table as an alternative to the developmental grids. Values for MCV are plotted in a similar manner. Again, there is close agreement among groups, especially at the fiftieth percentile. Values for boys and girls are virtually identical, except that the rise in MCV occurs at a somewhat earlier age in girls than in boys. There is greater variability at the third and ninety-seventh percentiles. Values for MCV rise very gradually through most of childhood, with a median value of 84 to 85 fl at 15 years of age that is still well below the adult value of about 90? Some data indicate that MCV values do not reach a plateau even in the adult, but that they continue to increase throughout life. ~~ 11 DISCUSSION There are two general approaches to establishing criteria of normality; one involves studying a random sample of the total population. This approach is exemplified by the recently published growth grids from the National Center for Health Statistics that provide a standard method for assessing growth of contemporary infants, children, and youth in the United States. 1..... One disadvantage &these grids is that they do not exclude values of abnormal individuals. For example, body weight curves reflect current conditions, with obesity a common abnormality. Inclusion of obese individuals in the derivation of weight grids may be misleading if the curves are interpreted as representing optimal standards. However, exclusion of obese individuals or other abnormal categories involves applying somewhat arbitrary definitions. Consequently, the use of height and weight values from a random sample of the population is the least controversial method of obtaining growth standards. A second group of methods for deriving normative data involves the exclusion of all subjects with evidence of
Percentile curves for hemoglobin and red cell volume
29
Table
Hemoglobin (gm/ dl) Age @r) 0.5-2 2-5 5-9 9-12 12-14 Female Male 14-18 Female Male
MCV (fl)
Median
Lower limit
Median
Lower limit
12.5 12.5 13.0 13.5
11,0 11.0 11.5 12.0
77 79 81 83
70 73 75 76
13.5 14.0
12.0 12.5
85 84
77 76
14.0 15.0
12.0 13.0
87 86
78 77
14.0 16.0
12.0 14.0
90 90
80 80
18-49
Female Male
common abnormalities or adverse environmental circumstances that might be expected to bias the data. Although this method involves the use of criteria that might seem arbitrary, it has certain advantages for the derivation of normative data for hemoglobin and mean corpuscular volume. Iron deficiency is particularly prevalent between the ages of I and 3, '~ and almost any sampling of this age group will include substantial numbers of children with iron deficiency anemia. The prevalence of thalassemia minor will vary according to the ethnic make-up of the population. Since normal standards are to be applied in screening for these two common conditions, it is desirable that the standards are calculated after excluding as many of these individuals as possible. The curves in the Figures represent an initial attempt to do this. Any criteria that reliably exclude iron deficiency also exclude some iron-sufficient subjects. Indeed, the more sensitive the criteria are in respect to excluding iron deficiency, the less specific they are likely to be by also excluding a larger percentage of the iron-sufficient population. Since it is likely that substantial numbers of iron-sufficient children were excluded by our criteria, it is important to ask whether the removal of this group biased the reference data. In the derivation of normal hemoglobin values, for example, it was assumed that hemoglobin concentration and MCV are dependent variables only when a condition such as iron deficiency or thalassemia minor results in both laboratory values being low. In the reference population, hemoglobin and MCV were considered to be independent variables. Although the latter is not strictly correct, it is close enough to the truth to be a reasonable working assumption. This was evident in US I, in which there was a very slight but significant tendency
30
Dallman and Siimes
for high MCV values to be associated with high hemoglobin concentration at any given age.' Based on normalized data from the entire group, we would predict that 10year-old white children, for example, with a higher than average MCV of 86 fl would have a mean hemoglobin of 13.5 gm/dl; with a lower than average MCV of 78 fl there would be a slightly lower hemoglobin of 13.4 gm/dl. Only when the MCV fell below the presumptive lower limit of normal of 76 (roughly corresponding to our exclusion criterion) was there a much steeper and more substantial fall in mean hemoglobin to 12.5 gm/dl. We felt it legitimate to assume that MCV and hemoglobin were independent variables within the reference population because the hemoglobin concentrations were unlikely to be altered by more than 0.1 gm/dl by excluding normal subjects on the basis of our MCV criteria. The same reasoning applies to the calculation of MCV after exclusion of low hemoglobin values. It could also be asked whether our criteria for transferfin saturation and serum ferritin would bias the calculation of reference values for hemoglobin and MCV. In iron deficiency, low transferrin saturation and serum ferritin concentrations are associated with low values of hemoglobin and MCV, but there is no evidence for a similar correlation in iron-sufficient individuals. We compared the mean serum iron in screened children of the three age groups in US III whose hemoglobin, MCV, and serum ferritin values, respectively, were below the median for each test with those whose values were above the median.' We found no significant difference in serum iron between the high and low hemoglobin, MCV, and serum ferritin values. US I was previously part of a study of hemoglobin values in whites, blacks, and Asians. As had been found in other surveys, 1:'-~ blacks were found to have consistently lower hemoglobin values by about 0.5 gm/dl.' This difference should probably be taken into account in using these grids. Data from a smaller number of Asians of various ethnic backgrounds (primarily Chinese, Japanese, and Filipino) were also analyzed, and there was no evidence that their values were different from those of whites after subjects with low MCV values were excluded. MCV values in blacks may also be somewhat lower than in whites? but the magnitude of this difference after exclusion of thalassemia minor and iron deficiency has not been determined. A comparison with earlier published values indicates that hemoglobin concentrations in this paper are higher than previous tabulations of normal values at many ages. Values at one year of age would be expected to be most subject to error if they are based on a population with a high prevalence of subclinical iron deficiency. In the large
The Journal of Pediatrics January 1979
study of Guest and Brown, '~ the mean hemoglobin concentration at 1 year of age was 11.1 gm/dl. Subsequent tabulations from textbooks and surveys range from 11.2 to 12.0 gm/dl}' .~0-24all substantially lower than our median value of 12.5 g m / d l ? Hunter and Smith z~ demonstrated that exclusion of subjects with laboratory evidence of iron deficiency substantially raised the distribution of hemoglobin and hematocrit values in a group of healthy infants of various racial groups. The fact that our values are derived from a white population is an additional explanation for the finding of higher values than in studies based on a more racially heterogeneous population. Alternative methods of deriving normative data for hemoglobin involve excluding iron-deficient subjects by employing a therapeutic trial of iron S. '-'~or assuming that hemoglobin values in healthy populations follow a single Gaussian distribution/'~ During adolescence, developmental changes in hemoglobin or hematocrit are a function of the stage of sexual maturation in boys. ~ It is possible to develop criteria of normal for each stage of puberty, but the use of percentile grids by age seems more practical for routine clinical use. Differences in onset of puberty account for the large spread between the third and ninety-seventh percentiles for hemoglobin in boys. For most laboratory studies, values are considered subnormal when they are more than 2 SD below the mean~ a lower limit that corresponds to the third percentile. In the case of hemoglobin, it is recognized that there is a large overlap between normal and subnormal values. ~..... For example, an individual may have a value that lies in the lower portion of the normal distribution for his sex and age despite having mild iron deficiency. This individual, whose hemoglobin or hematocrit would normally be higher, would not be recognized as being "anemic" unless he were treated with iron. '-'~By combining the MCV with the hemoglobin in screening for iron deficiency, '~ the accuracy of diagnosis is increased because it is less likely that both values would be in the lowest portion of the distribution curve by chance.' In populations in which iron deficiency is common, values for hemoglobin and MCV below the tenth percentile might be considered grounds for suspecting iron deficiency. However, in a population that contains few iron-deficient individuals, criteria should probably be more stringent because it is difficult to justify investigating too many normal children in the lower part of the normal distribution curve in order to discover every example of mild deficiency. The statistical considerations in evaluating these two types of populations are applicable to many screening situations? ~ The beneficial results of treating iron deficiency and counseling individuals with
Volume 94 Number 1
thalassemia m i n o r m u s t be b a l a n c e d against the cost a n d inconvenience of the screening p r o c e d u r e to n o r m a l subjects. Our d e v e l o p m e n t a l grids for h e m o g l o b i n a n d M C V are a n attempt to deal with currently available data. As more c o m p r e h e n s i v e survey data b e c o m e available, a n d with
Percentile curves for hemoglobin and red cell volume
14. 15.
16.
verification o f reference values by t h e r a p e u t i c trial with iron, it is likely that a d j u s t m e n t s a n d modifications will
17.
have to be made. REFERENCES
1. Dallman PR, Barr GD, Allen CM, and Shinefield HR: Hemoglobin concentration in white, black, and Oriental children: is there a need for separate criteria in screening for anemia? Am J Clin Nutr 31:377, 1978. 2. Frerichs RR, Webber LS, Srinivasan SR, and Berenson GS: Hemoglobin levels in children from a biracial southern community, Am J Public Health 67:841, 1977. 3. Dallman PR: Nutritional anemias, in Handbook of clinical nutrition, Evanston, American Academy of Pediatrics (in press). 4. Koerper MA, Mentzer WC, Brecher (3, and Dallman PR: Developmental change in red blood cell volume: Implications in screening infants and children for iron deficiency and thalassemia trait, J PEDIATR 89:580, 1976. 5. Koerper MA, and Dallman PR: Serum iron concentration and transferrin saturation in the diagnosis of iron deficiency in children: Normal developmental changes, J P~DIATR 91:870, 1977. 6. Saarinen UM, and Siimes MA: Developmental changes in serum iron, total iron-binding capacity and transferrin saturation in infancy, J PEDIATR 91:875, 1977. 7. Slimes MA, and Lundstr6m U: Unpublished data. 8. Saarinen UM, and Siimes MA: Developmental changes in the red blood cell counts and indices in infants with iron deficiency excluded both by laboratory criteria and by continuous iron supplementation, J PEDIATR 92:412, 1978. 9. Williams WJ, Beutler E, Erslev AL, and Rundles RW: Hematology, ed 2, 1977, McGraw-Hill Book Company, New York, p 10. 10. Okuno T: Red cell size as measured by the Coulter model S, J Clin Pathol 25:599, 1972. I1. Okuno T: Red cell size and age. Br Med J 1:569, 1972. 12. National Center for Health Statistics: NCHS growth charts, 1976, Monthly Vital Statistics Report, Vol 25, No 3, Suppl (HRA) 76-1120. 13. National Center for Health Statistics: NCHS growth charts, 1976, Vital and Health Statistics, Series 11, Health Resources Administration, Rockville, Md., U.S. Government Printing Office.
18.
19.
20. 21.
22.
23. 24.
25.
26. 27. 28.
29. 30. 31.
31
Lahey ME: Iron deficiency anemia, Pediatr Clin North Am 4:481, 1957. Ten-State Nutrition Survey, U.S. Department of Health, Education and Welfare, DHEW Publ. No. (HSM) 72-8130, 1972. Owen GM, Lubin AH, and Garry PJ: Hemoglobin levels according to age, race, and transferrin saturation in preschool children of comparable socioeconomic status, J PEDIATR 82:850, 1973. Garn SM, Smith NJ, and Clark DC: Lifelong differences in hemoglobin levels between blacks and whites, J Natl Med Assoc 67:91, 1975. Owen GM, and Yanochik-Owen A: Should there be a different definition of anemia in black and white children? Am J Public Health 67:865, 1977. Guest GM, and Brown EW: Erythrocytes and hemoglobin of the blood in infancy and childhood, III. Factors in variability, statistical studies, Am J Dis Child 93:486, 1957. Moe PJ: Normal red blood picture during the first 3 years of life, Acta Pediatr Scand 54:69, 1965. Burman D: Haemoglobin levels in normal infants aged 3 to 24 months and the effect of iron, Arch Dis Child 47:261, 1972. Hunter RE, and Smith N J: Hemoglobin and hematocrit values in iron deficiency in infancy, J PEDIATR 81:710, 1972. Wintrobe MM: Clinical hematology, ed 7, Philadelphia, 1974, Lea & Febiger, Publishers, p 1797. Dallman PR: Blood and blood forming tissues, in Rudolph AM, editor: Pediatrics, ed 16, Appleton-Century-Crofts, Inc, New York, 1977, p 1111. Garby L, Irnell L, and Werner I: Iron deficiency in women of fertile age in a Swedish community. II. Efficiency of several laboratory tests to predict the response to iron supplementation, Acta Med Scand 185:107, 1969. Cook JD, et al: Nutritional deficiency and anemia in Latin America, Blood 38:591, 1971. Daniel WA: Hematocrit: maturity relationship in adolescence, Pediatrics 52:388, 1973. Garby L, Irnell L, and Werner I: Iron deficiency in a Swedish community. III. Estimation of prevalence based on response to iron supplementation, Acta Med Scand 185:113, 1969. Cook JD, Finch CA, and Smith N J: Evaluation of the iron status of a population, Blood 48:449, 1976. Dallman PR: New approaches to screening for iron deficiency. J PEDIATR 90:678, 1977. Galen RS, and Gambino SR: Beyond normality: The predictive value and efficiency of medical diagnoses, New York, 1975, John Wiley & Sons, Inc.
