98 Dungy/ Morgan/Heotis/ Branson/ Adams 14 qT wv tT T T T 13+ » Fig. 1. Age-specific values of Hb in Marshallese children. ——- = Mean. = --- = median: ...... = mean, with Hb 2 1 al , y 88 xn as ' t F T+ T exclusions defined by Dallman and Siimes [7]; ----- = same as ......, but without exclusion of any Hb values from subjects with Hb higher than mean. T 86} 4 B44 4 got 4 80} 4 = val > 4 sions defined by Dallman and Siimes [7]; ----- = sameas. ......, but without ex- 4 clusion of MCV values from subjects 4 with Hb higher than mean. dian; ...... = mean, with MCV exciu- $ 76 & Fig. 2. Age-specific values for mean corpuscular volume in Marshal= Mean; --- = melese children. — 2 . 4 \ 6 , 8 i 10 — 12 14 Age, years were found to be 8.9 and 4.5%, respectively (table ITT). In estimating the prevalence of iron deficiency it was found that 12 of 412 children tested (2.9%) had an FEP which, by laboratory criteria, was ‘definitely’ elevated (> 60 ug/dl). The MCV was below the limits for normal set by Dallman and Siimes(7] for 21 of the 538 children tested (3.9%). Discussion Whatis the best way to determine the normal range of Hb in a population of infants and children? Which values should be included in the analysis and which values excluded? The reported normal ranges for Hb tend to be similar even though the exclusion limits may be derived differently. The lower limit for normal Hb in children aged 3-5 years has been placed at 11.0, 11.0, 10.9 or 11.2 g/dl using 4 different approaches [7, 9-11}. Nevertheless, it is important to 5012548 minimize systematic bias [12]. The distribution of Hb values in a normal population is known to be Gaussian, or nearly so [13], but abnormal values need not be so distributed. In populations where diseases associated with anemia are prominent, skewing of the total population’s Hb distribution will be toward lower levels. Smoking will skew values upward. Mean values will be shifted similarly. Therefore, it is an oversimplification to exclude high and low values of equivalent deviation from the mean (or median) if clinical experience suggests one pattern of deviation greatly predominates. Hb levels above the mean (or median) should be included in derivation of the normal range of Hb in children since a pathologically elevated Hb level in childhood is an infrequent occurrence and iron deficiency is common. Using this approach, the present analysis detected a small but systematic bias in the method used by Dallman and Siimes [7] which is apparent at every year for ages 1-13 (fig. 1). Because of the evidence of systematic