The relationship of race, sex, and age to concentrations of serum immunoglobulins expressed in international units in healthy adults in the USA*

BIOLOGICAL STANDARDIZATION The relationship of race, sex, and age to concentrations of serum immunoglobulins expressed in international units in healthy adults in the USA* S. E. MADDISON,1 C. C. STEWART,2 C. E. FARSHY,3 & C. B. REIMER4 Only afew investigations have been made to obtain human serum immunoglobulin values in units compatible with those used by the WHO International Reference Preparation for the Human Immunoglobulins IgG, IgA, and IgM. We report our summary statistics of serum IgG, IgA, and IgM, in international units (IU), for some 800 healthy American adults grouped by age, sex, and race. Our findings are in general agreement with some, but not with all, published data. We found that the mean IgG concentration is markedly higher and the mean IgA concentration is slightly higher in blacks than in whites. Except for white females, there was a significant increase in mean IgA with agefor both races. In the younger adults of both races, mean IgM values were markedly higher in females than in males. Statistically significant interactions between race, age, and sex factors were seen for all three immunoglobulin classes. Although we have attempted to estimate the normal population means and variances for the serum concentration of IgG, IgA, and IgM the process we used to select specimens may have resulted in some bias; much larger, truly randomized, and fully documented studies in different geographic areas and in different socioeconomic and racial groups are needed to provide accurate acceptable limits for human immunoglobulins. Serum immunoglobulin (Ig) determinations are used extensively, both as diagnostic indicators of disease and to monitor therapy in a variety of hypo- and hypergammaglobulinaemic states. Concentra- tions of the immunoglobulins have commonly been expressed in terms of mass concentration, such as mg/ml. However, comparable results between labo- ratories were not obtained until a common reference standard was used (21). The WHO International Reference Preparation of Human Immunoglobulins IgG, IgA, and IgM is available as a primary stan- dard. Each vial contains 100 international units (IU) each of IgG, IgA, and IgM. The problems resulting * From the Bureau of Laboratories, Center for Disease Control, US Public Health Service, US Department of Health, Education, and Welfare, Atlanta, GA 30333, USA. 1 Chief, Parasitic Immunochemistry Branch. 2Mathematical Statistician. ' Microbiologist. ' Chief, Immunological Products Branch. from the use of the unfamiliar IU, however, are well recognized and have resulted in the authorization (12) of conversion factors obtained from published data to relate the two systems of measurement: 80.4 ,ug of IgG, 14.2 ,tg of IgA, and 8.47 jug of IgM are equivalent to 1.00 IU. Only a limited number of investigations have been made to obtain values related to the WHO prepara- tion. An extensive investigation was carried out by Buckley & Dorsey (1) who determined the IgG, IgA, and IgM serum levels in IU in sera of 800 subjects ranging in age from 1 to 92 years. Buckley et al. (1, 2) developed a biomathematical model to aid the study of the effects of age, sex, and race on immuno- globulin levels. Rowe (22) has reported the IgG, IgA, and IgM serum concentrations in IU for healthy young adult males from 11 different coun- tries. Cejka et al. (4) reported similar data in IU for cord sera, and sera from some normal children and a few normal adults. 3337 - 179 - BULL. WORLD HEALTH ORGAN., Vol. 52, 1975 180 S. E. MADDISON ET AL. Numerous reports have indicated that statistically significant differences in the levels of serum IgG, IgA, and IgM are related to race, age, or sex. Although many of these studies are undoubtedly consistent internally, comparison between the values reported are uncertain because a common reference preparation was not used. Accuracy as well as precision is required if a diagnostic decision is to be made concerning the " normalcy " of a given serum sample. Here we report our findings of baseline data for serum IgG, IgA, and IgM in IU for some US adults grouped by age, sex, and race. MATERIALS AND METHODS All sera were obtained from the Center for Dis- ease Control serum bank. The specimens had been collected from healthy people between 1959 and 1969 during epidemiological surveys for specific anti- bodies following viral epidemics in the states of Alabama, Florida, Illinois, Louisiana, Mississippi, and Texas. The sera were stored at -20°C and were not thawed more than twice during the investigation. The study was initially designed to involve a total of 600 normal sera, with 150 from each of the following categories: white male, white female, black male, and black female. Within each of these cate- gories, the plan was to obtain 50 sera from indi- viduals in each of the following age groups: 20-40 years, 41-60 years, and 61-80 years. After the initial study had been completed we decided that more reliable estimates of population parameters were needed for IgA and IgM levels of white males between 61 and 80 years of age and for IgA levels of white females in the same age group. An additional 100 sera were therefore tested for each of these groups. Concentrations of IgG, IgA, and IgM were estimated by the single-radial-immunodiffusion tech- nique of Mancini et al. (15). Goat antisera prepared in our laboratories were shown to be monospecific by immunoelectrophoresis with whole human serum and also by inverse single-radial-immunodiffusion with purified Ig. Fifteen antigen wells in each plate were charged with appropriate dilutions of sera or reference pre- parations by using a 10-,ud Eppendorf a pipette. Four a Use of trade names is for identification only and does not constitute endorsement by the Public Health Service or by the US Department of Health, Education, and Welfare. wells contained dilutions of a secondary standard b that we had previously calibrated (in IU per ml) after a direct comparison with the International Reference Preparation: 67/97. At least one but usually two wells on each plate contained control samples (1 and 2) from each of two distinct pools of normal human serum. The remaining nine wells on each plate contained appropriate dilutions of un- knowns. Diffusion was allowed to proceed in a humid chamber at 200C for 4 days, followed by overnight drying and then staining of the plates. The diameters of precipitation disks (generally 6-9 mm) were measured to the nearest 0.1 mm. A calibration curve was constructed for each plate by using the least squares technique (19) to fit a linear line relationship between concentrations (in IU) of the dilution of the secondary standard and the corre- sponding areas of precipitate calculated from the measured diameter. These calibration curves were then used to estimate concentrations of all other samples contained on each plate. Further studies, including cellulose acetate electro- phoresis, immunoelectrophoresis, and gel filtration chromatography on Sephadex G 200, were carried out on sera with high IgG, IgA, or IgM levels. RESULTS Estimates of imprecision (lack of reproducibility) in our procedures for quantitating immunoglobulins were obtained from the values observed on two controls included in each plate. The coefficients of variation found were: for IgG, 10.9% and 11.5%; for IgA, 7.8% and 7.2%; and for IgM, 7.4% and 5.7%. Estimates of inaccuracy (lack of agreement with an accepted standard) in our procedure for quanti- tating immunoglobulins were obtained by compar- ing our results with results reported by others using coded (identical) serum samples in a completely blind performance evaluation of laboratories con- ducted by the Licensure and Proficiency Testing Division of the Bureau of Laboratories (27). Only 10 of the 125 laboratories participating in this study reported their results in IU. Results from these 10 laboratories gave coefficients of variation of 10.3% b The value assigned to the secondary standard is the mean of at least 29 determinations with some determinations performed on each of at least 6 different plates. The standard error is no greater than 1 IU for any assigned values; that is, one can state with 95% confidence that the limiting mean value is within 2 IU (twice the standard error) of the assigned value. SERUM IMMUNOGLOBULIN CONCENTRATIONS Table 1. Average values in International Units per ml of Descriptive statistics for the concentrations of coded control sera used in a completely blind evaluation IgG, IgA, and IgM (in IU per ml) for the samples accuracy among 10 participating referee laboratories studied are listed in Table 2 according to race, sex, Mean for 10 and age. The distribution of values showed some- Immuno- participating SD of mean Coefficient CDC mean b what greater spread for IgA and IgM than for IgG. globulin of variation class (u /m (%) (lU/m) Both very low and very high levels of the three immunoglobulins are found in all groups. Distribu- lgG 143.5 14.8 10.3 146.2 tions for all categories were skewed towards higher IgA 91.4 11.0 12.0 87.3 1gM 218.8 32.5 14.8 212.3 a Average of duplicate determinations. a Average of 3 determinations. values. Results from a three-factor analysis of variance technique (5, 8), used to investigate joint effects of race, sex, and age on the mean immunoglobulin concentrations, are shown in Table 3. A separate analysis was performed for each immunoglobulin for IgG, 12.0% for IgA, and 14.8% for IgM. Our class. results were in good agreement with the mean results We found that the mean IgG concentration is reported by these 10 laboratories (Table 1). markedly higher in blacks than in whites (P <0.01). Table 2. Median and mean concentrations and confidence intervals of the mean, in IU per ml,a for IgG, IgA, and IgM according to race, age, and sex White male White female Black male Black female 20-40 41-60 61-80 20-40 41-60 61-80 20-40 41-60 61-80 20-40 41-60 61-80 IgG No. of subjects median mean standard deviation 5th empirical percentile 95th empirical percentile IgA No. of subjects median mean standard deviation 5th empirical percentile 95th empirical percentile 1gM No. of subjects median mean standard deviation 5th empirical percentile 95th empirical percentile 50 50 49 122 135 123 119 135 126 25 36 40 72 78 60 166 204 215 50 50 146 111 143 147 117 145 157 42 61 69 59 56 64 209 271 294 50 50 148 139 107 128 152 129 144 68 66 90 58 37 44 299 279 321 49 50 50 131 126 112 131 132 121 30 33 41 68 81 68 183 202 204 50 50 148 99 139 119 117 143 130 51 71 61 59 44 47 231 264 244 49 50 50 212 200 120 220 207 130 95 114 77 75 60 34 455 475 284 49 47 49 162 186 186 168 192 187 43 40 39 100 126 124 250 270 258 47 47 49 134 147 195 147 164 210 61 87 95 53 45 82 274 383 432 50 47 49 145 128 92 169 141 118 75 69 63 79 40 44 321 303 244 49 50 48 186 178 185 192 180 187 53 40 46 112 114 126 298 263 282 50 50 49 141 168 186 143 178 199 46 71 70 76 72 94 230 310 348 50 50 49 246 132 98 256 152 137 95 80 108 110 54 49 456 333 477 aTo convert to mg/litre multiply these values by 0.804 for IgG, 0.142 for IgA, and 0.0847 for IgM; to convert to mg/dl multiply by 8.04,1.42, and 0.847, respectively. (See ref. 5.) 181 S. E. MADDISON ET AL. Table 3. Summary of analysis of variance results IgG IgA IgM Source of variation degrees degrees degrees of F-ratio of F-ratio of F-ratio freedom freedom freedom Race 1 303.94 a 1 55.33 a 1 0.07 Sex 1 0.66 1 0.98 1 36.93 a Age 2 1.65 2 24.32 a 2 33.24 a Race by sex 1 0.30 1 0.76 1 0.12 Race by age 2 1.03 2 5.08 a 2 4.20 b Sex by age 2 5.51 a 2 2.31 2 10.13 a Race by sex by age 2 0.85 2 0.32 2 5.02 a Within cells 578 774 680 a p = < o.o1. b p = < 0.05. One cannot conclude that the mean IgG levels are affected by age if sex is ignored, or by sex if age is ignored. However, depending on how the data are grouped, one can conclude that the change in IgG concentration with age is somewhat different in the two sexes, indicating that a statistically significant (P <0.01) interaction exists between sex and age factors independent of race. Mean IgA levels were also significantly higher in blacks than in whites (P<0.01). Except for white females, there was a significant (P <0.01) increase in mean IgA with age for both races; the exception indicates the presence of a significant interaction between race and age factors for IgA in white females. In the younger adults (20(40 years) of both races, mean IgM values were markedly higher (P <0.01) in females than in males. Statistically significant inter- actions between race, age, and sex factors for IgM (Table 3) were seen as a pronounced decrease in IgM levels with age, particularly in females of both races; the decrease was less pronounced in black males and did not occur at all in white males. Further tests to determine if high immunoglobulin levels observed in some sera reflected monoclonal proteins were carried out on 29 sera; in 22 of these the abnormally high immunoglobulins were shown to be polyclonal (IgA: 8; IgM: 11; IgG: 3). These data are included in the results of Table 2. The remaining 7 proteins studied were monoclonal. Data from these sera were not included in Table 2. DISCUSSION The importance of the WHO International Refer- ence Preparation of Human Immunoglobulins IgG, IgA, and IgM lies in its potential diagnostic useful- ness for comparing observations carefully made at any time in any laboratory on the serum of any individual patient with observations of a normal population, once this baseline data in IU is well established and made generally available. For valid comparison both precision and accuracy are re- quired. Both the methodology and the reagents used can influence final results. Obviously, results will be biased if the antisera used are not specific, particu- larly if they react with light chains to bind other Ig classes into the same precipitation lattice with the Ig being measured. Nonspecificity of this type has been found in commercial kits sold for diagnostic use (20). Subclass selectivity of the antisera may also bias the results. The quality of the secondary standard calibrated in terms of the WHO preparation is of utmost importance. Purity and retention of the antigenic and conformational characteristics of the native molecule are essential. Methodologically, imprecision when filling wells can be a major source of error (7). By using an Eppendorf pipette we believe that our error from this source is less than 1.0 %. Following the essential characteristics of the original Mancini procedure (15), we have allowed 4 days at room temperature for the precipitation reaction to reach quasi-equilib- rium instead of using the kinetic modification of Fahey & McKelvey (6). Recently, Heremans et al. (10) have emphasized that the kinetic measurements can only be made at the expense of accuracy and information. As assessed in our laboratory by the coefficient of 182 SERUM IMMUNOGLOBULIN CONCENTRATIONS variation of internal controls included in every radial diffusion plate, our imprecision of measurement was much less than the spread of Ig values in healthy populations. Our reproducibility is similar to that seported by others (1, 13). The higher coefficient of variation of our IgG determinations in comparison with our IgA and IgM determinations probably reflects the two dilution steps we used to obtain an appropriate dilution for IgG as opposed to the single dilution step used for the IgA and IgM determina- tions. The sera we studied had been stored at -20°C for 4-14 years before testing. Veys & Wieme (28) reported that IgG levels were not influenced by relatively short-term storage of samples at -20°C, but that IgA and IgM concentrations declined when sera were stored for 3 months. In the present study the long period of storage at -20°C before testing may have affected the immunoglobulin levels re- ported here, but we have no evidence of this. We consider it unlikely that the storage contributed to or detracted from the statistically significant differences we observed in the immunoglobulin concentrations within the various groups. We have not detected significant changes in IgM levels following repetitive freezing and thawing of sera (unpublished data). The number of monoclonal proteins detected here (0.75% in 800) may well be an underestimate, be- cause IgG values were not determined on 200 sera from the older age groups. Carrell et al. (3) observed 0.5% of monoclonal proteins in the population of 2 192 adults in a New Zealand town. The occurrence of the monoclonal gammopathies in our study in the elder age groups agrees with previous findings. Previous studies of levels of immunoglobulins in normal adults, for the most part reported in milli- gram values or in percentages of normal adult levels, have shown variation in Ig levels with race, age, and sex. This line of investigation has been comprehen- sively reviewed by Hobbs (11). In trying to establish normal adult values for immunoglobulin levels in terms of IU we found, as have others, that race, age, and sex have a significant effect on the mean values of IgG, IgA, and IgM in the various groups of individuals. Although the published variations re- ported are not in complete accord, they do stress the importance of taking these variables into account. Grundbacher (9) has presented evidence that in humans the X chromosome carries genes with an effect on IgM levels, thus accounting for the elevated levels observed in younger adult females, both black and white, compared with the IgM levels of males of the same age group. In contrast, Stiehm & Fuden- berg (25) reported no racial differences. Buckley & Dorsey (1) reported that white males had less IgG and IgA than black males, but IgM concentrations in these two groups were not significantly different. Karayalcin et al. (14) observed that blacks, both male and female, had higher levels of IgG, IgA, and IgM than whites. Our studies were carried out on a larger number of samples than were used in most reported investi- gations, and the sera in our study were drawn from surveys of healthy populations fairly representative of a cross-section of each area sampled. We recog- nize that the sample was not randomly selected and that this may result in some bias in our findings. Unfortunately, in no study yet published has the population sample been fully documented and ran- domly selected. In a number of previous studies the subjects investigated were drawn partly or entirely from hospital personnel. The size as well as the heterogeneity of our population sample may account for the wide range of immunoglobulin concentra- tions observed in all the groups we studied. Our data seem to be in general agreement with that reported in IU for adults by Rowe (22), Sinkov et al. (24), and Cejka et al. (4), and with IU data reported by 10 reference laboratories participating in a completely blind laboratory proficiency testing study conducted by the US Center for Disease Control (27) (Table 1). The basis of disagreement between our data and the IU data of Buckley et al. (1, 2) is unknown, but we note that the conversion factors suggested by Buckley & Dorsey (1) are quite different from those recently recommended by Humphrey & Batty (12). In attempting a comparative analysis of the results of previous studies of immunoglobulin levels carried out in various geographic areas and in different racial groups (1, 14, 16, 17, 18, 22, 23, 26, 29, 30) it becomes apparent that sex, race, age, climate, nutri- tion, and possibly as yet unrecognized factors may influence serum immunoglobulin levels. Although we have attempted to estimate accurately the normal US human population means and variance for the serum concentrations of IgG, IgA, and IgM, a much larger, randomly selected, and well docu- mented population study is needed to provide true normative limits. 3 183 184 S. E. MADDISON ET AL. ACKNOWLEDGEMENTS We thank our CDC co-workers for their generous help: Mr Charles F. Peters for supplying all of the sera, Miss Karen M. Fulford for information concerning proficiency testing results, Miss Diane Mundhenk for performing some of the assays, and Mrs Mary B. Felker for statistical evaluations. RtSUMt LA RELATION ENTRE LA RACE, LE SEXE ET L 'AGE ET LES CONCENTRATIONS DIIMMUNOGLOBULINES SfiRIQUES EXPRIMtES EN UNITES INTERNATIONALES CHEZ DES ADULTES EN BONNE SANTE AUX E'TATS-UNIS D'AME'RIQUE L'emploi de la preparation intemationale de reference OMS pour les concentrations d'immunoglobulines a per- mis aux laboratoires d'effectuer des determinations com- parables. Toutefois, rares sont les recherches qui ont ete faites en vue d'obtenir des titres d'immunoglobuline serique humaine en unites susceptibles d'etre exprimees dans les unit6s internationales (UI) utilisees pour les titres designes de IgG, IgA et IgM dans la preparation de reference de l'OMS. Cet article contient une recapitula- tion statistique concernant les IgG, IgA et IgM seriques exprimees en UI pour quelque 800 adultes en bonne sante residant aux Etats-Unis groupes par Age, sexe et race. Les specimens ont ete recueillis chez des gens en bonne sante entre 1959 et 1969 A l'occasion de prospections epidemiologiques effectu6es pour trouver des anticorps specifiques A la suite d'epidemies virales dans les Etats de l'Alabama, de la Floride, de l'Illinois, de la Louisiane, du Mississipi et du Texas, et ils ont ete stockes A -20'C. De meme que pour les autres etudes de titres d'immuno- globuline dont les resultats ont ete publies, l'6chantillon de population n'a pas fait l'objet d'un sondage al6atoire, de sorte qu'il se peut qu'il reflete quelque erreur systema- tique meconnue. Dans l'ensemble, nos conclusions concordent avec quelques-unes des donn6es - mais non la totalite - communiquees en UI pour les adultes, ainsi qu'avec les donn&es en UI signalees par dix laboratoires de reference participant A une etude effectu&e par le Center for Disease Control des Etats-Unis en vue de verifier la qualite du travail dans les laboratoires. La concentration moyenne d'IgG etait nettement plus elevee chez les noirs que chez les blancs (P < 0,01). On ne peut conclure que les niveaux moyens d'IgG sont affectes par l'Age si l'on ne tient pas compte du sexe, ni par le sexe si l'on ne tient pas compte de l'Age. Cependant, selon la faron dont les donnees sont groupees, on peut conclure que la modification de la concentration d'IgG avec l'Age differe quelque peu chez les deux sexes, ce qui indique qu'il existe une interaction statistiquement appre- ciable (P <0,01) entre le facteur sexe et le facteur age independamment de la race. Les niveaux moyens d'IgA etaient, eux aussi, nette- ment plus elev6s chez les noirs que chez les blancs (P < 0,01). Sauf chez les femmes blanches, I'IgA moyenne augmentait de fa9on appreciable (P <0,01) avec l'age chez les deux races; l'exception indique la presence d'une interaction notable entre le facteur race et le facteur age pour l'IgA chez les femmes blanches. Chez les adultes plus jeunes (de 20 A 40 ans) des deux races, les titres moyens d'IgM 6taient nettement plus elev6s (P < 0,01) chez les femmes que chez les hommes. Des interactions statistiquement appreciables entre les facteurs race, age et sexe pour l'IgM ont et6 observ6es comme correspondant A une diminution prononc6e des titres d'IgM avec l'age, en particulier chez les femmes des deux races; la diminution etait moins accentuee chez les hommes de race noire et nulle chez les hommes de race blanche. Bien que nous ayons cherche A obtenir une estimation des moyennes et des variations normales chez la popula- tion des Etats-Unis pour les concentrations d'IgG, d'IgA et d'IgM seriques, il est necessaire, pour obtenir les limites exactes, de proceder a une etude beaucoup plus vaste, veritablement fondee sur un sondage aleatoire, et parfaitement documentee. I1 ressort des enquetes ante- rieures qu'il faudrait effectuer des etudes analogues dans differentes regions du monde et chez differents groupes socio-6conomiques et raciaux afin d'6valuer complete- ment l'influence de ces facteurs sur les titres d'immuno- globuline. REFERENCES 1. BucKLEY, C. E. & DORSEY, F. C. Annals of internal medicine, 75: 673 (1971). 2. BUCKLEY, C. E. ET AL. Federation proceedings, 33: 2036 (1974). SERUM IMMUNOGLOBULIN CONCENTRATIONS 185 3. CARRELL, R. W. ET AL. Australian and New Zealand journal of medicine, 1: 398 (1971). 4. CEJKA, J. ET AL. Clinical chemistry, 20: 656 (1974). 5. DIXON, W. J. B.M.D. biomedical computer program. Berkeley, University of California Press, 1973. 6. FAHEY, J. L. & MCKELVEY, E. M. Journal of immunology, 94: 84 (1965). 7. FERGUSON, P. ET AL. Scottish medicaljournal, 19: 113 (1974). 8. GRAYBILL, F. A. An introduction to linear statistical models, Vol. I. New York, McGraw-Hill, 1961. 9. GRUNDBACHER, F. J. Science, 176: 311 (1972). 10. HEREMANS, J. F. & MASSON, P. L. Clinical chemistry, 19: 294 (1973). 11. HOBBS, J. R. Advances in clinical chemistry, 14: 291 (1971). 12. HUMPHREY, J. H. & BATTY, I. Clinical and experi- mental immunology, 17: 708 (1974). 13. KALFF, M. W. Clinical biochemistry, 3: 91 (1970). 14. KARAYALCIN, G. ET AL. New York State journal of medicine, 73: 751 (1973). 15. MANCINI, G. ET AL. Immunochemistry, 2: 235 (1965). 16. MCFARLANE, H. Lancet, 2: 445 (1966). 17. MCGREGOR, I. A. ET AL. Clinical and experimental immunology, 7: 51 (1970). 18. MOHAMMED, I. ET AL. Lancet, 1: 481 (1973). 19. OSTLE, B. Statistics in research, 2nd ed. Ames, Iowa State University Press, 1963, p. 161. 20. REMER, C. B. Health laboratory science, 9: 178 (1973). 21. ROWE, D. S. ET AL. Bulletin of the World Health Organization, 42: 535 (1970). 22. ROWE, D. S. Lancet, 2: 1232 (1972). 23. SAMUEL, A. M. ET AL. Indian journal of medical research, 58: 56 (1970). 24. SrNKOv, D. ET AL. Bulletin of the World Health Organization, 49: 217 (1973). 25. STIEHM, E. R. & FUDENBERG, H. H. Pediatrics, 37: 715 (1966). 26. TURNER, M. W. & VOLLER, A. Journal of tropical medicine and hygiene, 69: 99 (1966). 27. US DEPARTMENT OF HEALTH, EDUCATION AND WELFARE. Proficiency testing: non-syphilis serology, quantitative immunoglobulins. Atlanta, Center for Disease Control, 1973. 28. VEYS, E. M. & WIEME, R. J. Clinica chimica acta, 47: 295 (1973). 29. YADAV, M. & SHAH, F. H. Lancet, 2: 450 (1973). 30. ZEGERS, B. J. M. ET AL. Vox sanguinis, 24: 457 (1973).