Original Research Communications-surveys Diagnosis of anemia and iron deficiency: analytic and biological variations of laboratory tests13 Peter R Daliman, MD ABSTRACT We describe the magnitude of analytic errors and the within-subject biological variations for laboratory tests that are used in the diagnosis of anemia and iron deficiency. For Hb, hematocrit, and red cell indices, coefficients of variation for analytic and biological variations are less than 4%. In general, higher coefficients of variation are characteristic of serum iron, ironbinding capacity, serum ferritin, and erythrocyte protoporphyrin. Particularly high analytic variations between different laboratories have been described for iron-binding capacity and the greatest biological variations have been noted for serum iron. An awareness of these sources of error is helpful in designing studies and in interpreting laboratory results. Am J C/in Nuir 1984;39:937-941. KEY WORDS Anemia, iron deficiency, analytic variations, biological variations, diagnosis Introduction During the past 20 yr, the focus of attention in the diagnosis of anemia and iron deficiency has shifted from the severe cases that are usually seen in hospitals to the more common mild cases that are typically seen in an outpatient setting. Concurrently, an increasing number of laboratory tests has become available, many tests have been automated, quality control has been more widely applied, and reference standards according to age and sex have become more precisely defined ( 1-3). The general improvement in laboratory tests has been a necessary development in the diagnosis of mild anemia and iron deficiency because laboratory results deviate less from the normal range than they do in severe cases; indeed they often overlap into the normal range (4, 5). In designing studies or in evaluating patients, it therefore becomes more important than ever to have some awareness of the analytic errors and the within-subject biological variations for each of the laboratory tests and some guidelines for deciding whether a change in results is actually related to treatment. In the last few years, a number ofstudies have provided information on analytic and biological variations for Hb, hematocrit (Hct), red cell indices, serum iron and iron-binding capacity, serum ferritin, and erythrocyte protoporphyrin. The purpose ofthis report is to provide a convenient summary of some of this information. Methods and statistical analyses Analytic and biological variations forlaboratory tests are expressed in terms ofthe coefficient ofvariation for repeated measurements on the same specimen or the same subject. The coefficient of variation represents the SD of the repeated measurements expressed as a percentage ofthe mean. For example, if a normal subject s Hb concentration on eight sequential weekly blood From the Department of Pediatrics, University of California School of Medicine, San Francisco, CA 94143. 2Supported by NIH Grant AM 13897-14. 3Address reprint requests to: Peter R Dallman, MD, Professor of Pediatrics, Department of Pediatrics, Room M-6S0, University of California at San Francisco, San Francisco, CA 94143. Received December 12, 1983. Accepted for publication January 31, 1984. The 4,nerican Journal of Clinical Nutrition 39: JUNE 1984, pp 937-941. Printed in USA c 1984 American Society for Clinical Nutrition 937
938 DALLMAN samples averaged 10.0 g/dl and had a SD of ±0.3 g/dl, the coefficient ofvariation would be 3.0%. Since 2 SDs encompass the 95% confidence limits, a repeat venous Hb would then be expected to be within about 0.6 g/di of the initial value 95% of the time. Conversely, a change ofgreater magnitude in Hb concentration would be more likely to represent a true alteration. The studies that are cited for the tabulation of coefficients of variation in Table I are based on venous blood, drawn from healthy subjects under carefully controlled conditions, and analyzed by laboratories with a special interest in quality control. Variations under ordinary clinical circumstances can be expected to be greater. When coefficients of variation were calculated at several concentrations, that value closest to the iron-deficient range was selected. In reporting biological variations over hours, we have included only the period between 8 AM and 4 PM, because most analyses are done within this time span. The studies that have been cited in Table 1 (6-18) fall primarily into two categories. One group of studies involved the distribution of quality control samples to between nine and 1100 cooperating laboratories (6, 8, II. 13, 16, 18). In another category of investigations, groups of healthy adults, consisting of nine to 20 students, employees, and/or investigators, had repeated blood samples drawn during a day, on sequential days, and/or on sequential weeks (7,9, 10, 12, 14). In another study. 20 replicate measurements were made on a single sample and serial analysis were done on a daily basis on 22 stored samples (15). The term biological variation, used herein, refers only to the variations in laboratory tests that occur in individual subjects. The degree ofvariability that occurs within groups of individuals is indicated by the 95% reference ranges or by the SD for a given laboratory test (1, 2). The laboratory tests that are used in the diagnosis of iron deficiency (and anemia) can be conveniently grouped into screening and confirmatory tests. The initial screening test usually consists of either the Hb concentration or the Hct. Hb is measured after accurate dilution of the blood specimen in a solution that converts Hb to cyanmethemoglobin, which is then quantitated spectrophotometrically. The analysis is done either with a separate spectrophotometer or with the spectrophotometer component ofan electronic counter. The electronic counter will also provide a calculated Hct; however, this determination is not considered as reliable a means of diagnosing anemia as is the Hb. In office and clinic laboratories, the Hct is often measured by centrifugation of a minute amount of blood that has been collected in a heparinized capillary tube. The Hct is then calculated by comparing the height of the column of packed red cells with the height of the entire column of red cells and plasma. The most useful red blood cell (RBC) indices are the mean corpuscular volume (MCV) and mean corpuscular Hb (MCH). These measurements were rarely in routine use until the development of electronic counters. The manual determination of MCV and MCH required the microscopic estimation of the RBC, TABLE I Analytic and biological variations in tests related to diagnosis of anemia and iron deficiency Hb Hct Coulter Centrifuged MCV, Coulter Fe Automated Manual TIBC SF EP Extraction Hematofluorometer * References are in parentheses. Withinassay/day 2.4 (6)* Analytic days 2.8(6) 1.3(8) 2.9 (6) 3.3 (6) 1.8 (8) 2.7 (8) 0.9(6) 1.8(6) 3.1 (9) 4.6(6) 4.4(10) 3.2(9) 3.7 (10) 9.6 (6) 1.4(9) 5.8(6) 2.0(13) 3.5(9) 3.2 (9) 7.2 (9) Variations Coefficient of variation (%) labs 3.1 (6) 4.0 (6) 2.6 (6) 5(11) 14(11) 5 (13) 1.8(15) 5.0(16) 9.3(16) 7 (17) 1.4(16) 7.8(16) 2.5(17) 7.0(18) Within day 2.5 (7)t Biological t Reference 7 used ANOVA to estimate biological variation independent of analytic variation. variation days/wk 2.4(7) 2.6 (7) 2.5 (7) 12.9 (12) 28.5 (9) 26.6 (12) 29.3 (10) 4.8(9) 8.8 (10) 6(14) 14.5(9)
ANALYTIC AND BIOLOGICAL VARIATIONS 939 a procedure with a large analytic error. MCV and RBC can be measured directly by electronic counter. MCH is derived by dividing the Hb by the RBC. Among the tests that can be termed confirmatory tests for iron deficiency are serum iron (Fe) and ironbinding capacity (TIBC), serum ferritin (SF), and erythrocyte protoporphyrin (EP). Fe and TIBC are generally measured by spectrophotometric techniques. The assay is time consuming and subject to substantial errors due to contamination by iron from the environment when done manually. Automated techniques make it possible not only to obtain results more rapidly but to achieve greater reproducibility. The transferrin saturation is calculated by dividing the concentration of serum iron by the TIBC and multiplying by 100 to express the results as a percentage. Thus, transferrin saturation will reflect the biological variability and laboratory errors of both the serum iron and TIBC. The iron-binding protein, transferrin, binds most of the iron in serum. There are several assays available for the measurement of transferrin, but they are not in widespread clinical use. For this reason, only TIBC is included in Table 1. SF is generally measured by radioimmunoassay and that is the method by which the results summarized herein were obtained. Enzyme-linked immunoassays are being developed, and these are increasing in popularity. Recently, an international standard preparation offerritin has become available and this should improve agreement among laboratories in the future. EP is most frequently measured by so called hematofluorometers, dedicated instruments that provide an immediate reading from the reflected fluorescence of a thin film of blood. This method probably requires standardization against quality control samples in which EP was measured by an extraction method. Extraction methods also involve the measurement of fluorescence and are more readily standardized. Results and discussion Table 1 summarizes the coefficients of variation that are related to the analysis itself (multiple analyses ofthe same sample within an assay, within the same day, or between many days) and those that include both analytic variations and biological variations (analyses of multiple samples taken from the same individual within a day or at daily or weekly intervals). The screening tests that are obtained by electronic counter, Hb, Hct, RBC (not shown in Table 1), MCV, MCH, and MCHC (not shown in Table 1), fall into a category that is characterized by both low analytic and biological variations. The analytic variations range between about 1 and 3% and biological variations are within the same range. Biological variations were not calculated for the RBC indices in the paper cited in Table 1, but low values can be inferred from the fact that the hematocrit are derived from both the MCV and RBC. The same considerations apply to the values for MCH (derived from Hb and RBC). The errors for Hb done separately by the cyanmethemoglobin method or Hct by centrifugation of a heparinized capillary tube can be similarly low. However, instrumentation can be a limiting factor, and much higher analytic errors may be anticipated, especially with less precise instruments designed for field use. Nevertheless, biological and analytic variations are small in relation to variations among healthy individuals. To provide a frame of reference, the 95% range of repeated Hb measurements in an individual adult can be expected to fall within ± 1.0 g/ dl. In contrast, the 95 % reference range for Hb in healthy adult males between 18 and 44 yr of age encompasses ±2.0 g/dl (19). One of the implications of the relatively low biological and analytic errors for Hb, Hct, and the RBC indices is that they warrant more careful attention to the reference standards. In the case of Hb, for example, the widely used WHO lower limits of normal were rounded out to the nearest g/dl. Since distinctions between Hb values are meaningful to the next decimal place, the more recent trend toward lower limits given to the nearest 0. 1 g/dl seems entirely appropriate. In general, the confirmatory tests are characterized by higher analytic and biological errors than the screening tests. In the case of Fe, analytic variations can be drastically decreased by the use of automated equipment because the opportunity for environmental contamination is diminished. Variations due to biological factors are much greater than analytic variations with an automated method. Variations due to diurnal factors can be minimized by sampling in the morning or early afternoon (values can normally fall to very low levels at night). However, variations in results obtained at the same time of the day on sequential days or weeks remain substantially greater than analytic variations, even when samples are taken after an overnight fast. These biological variations are so substantial that further reduction of analytic errors will accomplish relatively little in the evaluation of individual
940 DALLMAN patients. There is an impression that the biological variations in Fe diminish in iron deficiency, resulting in values that are more consistently low, but it is difficult to find documentation for this. Within laboratories, the analytic errors for TIBC are in the same general range as those for Fe. Higher between-laboratory variations may be due, in part, to differences in methods. However, biological variations are much lower than those for Fe and scarcely exceed the between-day analytic variations. Consequently, the biological variations in Fe/TIBC can be attributed primarily to fluctuations in Fe. In the case of SF, the pattern of variations resembles that of TIBC. The within-assay analytic coefficient of variation is about 3% and about double that figure between days. The between-laboratory variation represents more than an additional doubling. This is not surprising in view of the large number of different radioimmunoassays and immunoradiometric assays that were used. The biological variations in SF are greater than those for TIBC but substantially below those for Fe. Results for EP are shown for both the more widely used hematofluorometer methods and the more readily standardized extraction methods. The within-assay coefficient of variation for the extraction method is very low but more than doubles between assays. The between laboratory agreement is better than that for SF as might be anticipated when comparing an extraction method with a radioimmunoassay. The hematofluorometer results are surprisingly good, with lower between-day variations than the extraction method. -laboratory coefficients of variation were similar for the two methods. Conclusion Remarkably consistent results can now be obtained by experienced laboratories for Hb, Hct, and RBC indices. This facilitates the detection of mild anemia. Furthermore, the relatively small biological variations in these laboratory measurements make it easier to distinguish even a relatively small response to therapy from a random fluctuation. In the case of SF, TIBC, and EP, further improvements can be anticipated as methods become more uniform and better standardized. Fe represents a special case in that biological variations override analytic ones. Whether this also holds in iron deficiency is uncertain. With improvements in methodology, a major remaining problem will relate to the overlap of test results between mildly iron-deficient and normal subjects (4, 5). This overlap may make it difficult to identify individuals with mild iron deficiency but is not a major obstacle to characterizing grous of patients or subjects. References I. Dallman PR, Siimes MA, Stekel A. Iron deficiency in infancy and childhood. Am J Clin Nutr l980;33:86-l 18. 2. Bothwell TH, Charlton RW, Cook JD, Finch CA. Iron metabolism in man. London: Blackwell Scientific Publications, 1979. 3. Cook JD, ed. Methods in hematology: iron. New York, NY: Churchill Livingstone, 1980. 4. Cook JD, Finch CA, Smith NJ. Evaluation of the iron status ofa population. Blood l976;48:449-ss. S. Dailman PR, Reeves, JD, Driggers DA, Lo EYT. Diagnosis of iron deficiency: the limitations of laboratory tests in predicting response to iron treatment in 1-year-old infants. J Pediatr 198 l;99:376-81. 6. Kaplow LS. Schauble MK. Bechtel JM. Validity of hematologtc data in Veterans Administration Hospital laboratories. Am J Clin Pathol l979;7 1:291-300. 7. Statland BE, Winkel P. Harris SC, Burdsall Mi, Saunders AM. Evaluation of biologic sources of variation of leukocyte counts and other hematologic quantities using very precise automated analyzers. Am J Clin Pathol l977;69:48-s4. 8. ROss JW, Frazer MD, Moore TD. Analytic clinical laboratory precision-state of the art for 31 analysates. Am J CIin Pathol l980;74:s2 1-30. 9. Pilou VA, Howanitz IJ, Howanitz JH, Domres N. Day-to-day variation in serum ferritin concentration in healthy subjects. Clin Chem 198 l;27:78-82. 10. Statland BE, Winkel P. Relationship of day-to-day variation of serum iron concentration to iron-binding capacity in healthy young women. Am J Clin Pathol 1977;67:84-90. 11. Itano M. College of American Pathologists. Comprehensive chemistry survey bonus option: iron binding capacity. Am J Clin Pathol l976;66:244-7. 12. Statland BE, Winkel P, Bokelund H. Variation of serum iron concentration in healthy young men: within-day and day-to-day changes. Clin Biochem l976;9:26-9. 13. International Committee for Standardization in Hematology. The measurement of total and unsaturated iron binding capacity in serum. Br J Haematol l978;38:28l-90.
ANALYTIC AND BIOLOGICAL VARIATIONS 941 14. Dawkins 5, Cavill I, Ricketts C, Worwood M. Variability of serum ferritin concentration in normal subjects. Clin Lab Haematol l979;l:4l-6. 15. Chisholm JJ, Brown DH. Micro-scale photofluorometric determination of free erythrocyte protoporphyrin (protoporphyrin IX). Clin Chem 1975;2 1: 1669-82. 16. Erythrocyte protoporhyrin proficiency testing. Atlanta, GA: Centers for Disease Control, January, May, June, September, 1981. 17. Schifman RB, Finley PR. Measurement of nearnormal concentrations of erythrocyte protoporphyrin with the Hematofluorometer: influence of plasma in front surface illumination assay. Clin Chem 198l;27:lS3-6. 18. Jackson KW. Interlaboratory comparison of results of erythrocyte protoporphyrin analysis. Clin Chem 1978;24:2l 35-8. 19. Dallman PR, Yip R, Johnson CL. Prevalence and causes of anemia in the United States, 1976 to 1980. Am J Clin Nutr l984;39:437-45.