Geometric means, reference ranges, and selected percentile points for blood Volume measurements by age, race/ethnicity, and gender

Background: Blood volume measurements provide important clinical parameters in certain treatment situations. Method: While a fast, accurate and reliable methodology has been available to estimate the blood volume measurements for almost a decade, the availability of “normal” or “should be” blood volume measurements have not been readily available. In this article, we have provided geometric means and selected percentiles points with 95% confidence intervals for total blood volume, red cell volume, and plasma volume for children aged 8-11 years old, adolescents 12-17 years old, and adults aged 18 years and over by gender, race/ethnicity, age, and body mass index. The reference ranges are also provided. Result: We have also provided mathematical formulae that can be used to compute “should be” blood measurements for specific patients with a given gender, race/ethnicity, age, and body mass index. Conclusion: It is up to individual physicians to decide in a given treatment situation which “should be” index, e.g., geometric mean vs. median they want to use for comparison with actual estimated blood measurements.


Introduction
Total blood volume (TBV) consists of red cell volume (RCV) and plasma volume (PV).Traditionally, normal values for TBV have been expressed as ml/kg total body weight (TBW) [1] .Other normalizing constants used to express total blood volume are the body surface area (BSA) and body mass index (BMI) [2] .Based on 50 "normal" males and 50 "normal" females aged 18 years or greater, mean TBV/TBW for females was 55.1 ml/kg (SD = 8.88) and 73.3 ml/kg (SD = 16.4) for males [2] .Mean TBV/BSA for females was 2.44 L/m 2 (SD = 0.26) and 2.78 L/m 2 (SD = 0.28) for males.However, since, adipose tissue is relatively avascular [3]   , TBV depends up on the body composition, i.e., lean body mass (LBM) and adipose tissue mass (ATM).Neither TBW nor BSA accounts for differences in LBM and ATM from one person to another and as such, TBV measurements based on them are subject to misinterpretation.Use of TBW as the normalizing constant results in overestimation of TBV in obese individuals and underestimation in lean individuals [2] .Use of BSA results in underestimation of TBV in both underweight and obese individuals [2] .According to Feldschuh and Enson [2] , use of the deviation of the TBW from the ideal body weight is the preferred reference method

RESEARCH ARTICLE
for predicting TBV.In a study of 77 children aged 18 years and under, Raes et al. [4] used LBM as a normalizing constant for TBV and found it to be the best index of normalization from among TBW, BSA, LBM, and BMI, and it was not affected by gender differences.However, LBM is not always easy to measure and the equations used by them [4] to compute LBM may not always be applicable.None of the studies reviewed above were based on large populations and as such do not provide reference values and ranges that can be used for general populations, like general U. S. population.In addition, none of them has looked at the issue of racial/ethnic differences in TBV if they do exist.However, based on the recommendations of an expert panel on rdionuclides appointed by the International Council of Standardization in Hematology [3] , the following prediction equations can be used to compute TBV which can then used to develop reference ranges for TBV.where BSA = TBW 0.425 * H 0.725 * 0.007184; TBW is measured in kilograms and height H, is measured in centimeters.And, in each case, TBV = RCV + PV.
An accurate and bias-free method of measuring and predicting normal blood volume can be a useful tool in managing a variety of disease states [2] .For example, hemodilution or increased plasma volume may be related to poor outcome in anemic chronic heart failure patients and patients with hemodilution may do worse than those with true anemia, i.e. reduction in red blood cells without hemodilution [5] .Accurate TBV measurements may affect treatment decisions [6] in surgical intensive care unit patients.Hypertensive patients have diminished TBV [7] .
However, in spite of the importance of the accurate measurement of TBV, measurement of TBV has not been a favorite beyond some hematologists until recently.For years, while a dual-isotope/dual-tracer technique which does the simultaneous measurements of RCV and PV has been a gold standard, the procure does not provide timely measurements -it takes about 4-6 hours or more to complete the procedureand presents many opportunities for error and the procedure is complex, for example, in one of the steps, it requires withdrawal and re-infusion of patients' own blood [8] .However, since its approval by U. S. Food and Drug Administration in 1998, a semiautomatic blood volume analyzer BVA-100, based on indicator dilution principle has solved some of the issues with the original gold standard [8] .This procedure provides results within about 90 minutes [8] .The return of blood volume as a clinical parameter has been welcomed [9] .
For the blood volume analysis to be useful in clinical practice, it should be able to answer two questions accurately and reliably.In other words, "what is" the TBV and what "should be" the TBV for a given patient under clinical care.While the answer to "what the TBV is" for a given patient is likely to be provided by the semiautomatic blood volume analyzer BVA-100, the answer to "what the TBV should be" does not seem to have readily available answer.It is our intent, in this study, to make available adequate information that will provide answer to this question.Specifically, we will generate tables for RCV, PV, and TBV with their geometric means, and selected percentile points each with 95% confidence intervals; and reference ranges computed as the difference between 2.5 th and 97.5 th percentiles points.Similar tables for TBV, PV, and RCV normalized by TBW will also be provided as supplemental material.We will also provide prediction equations that can be used to compute "should be" TBV, PV, and RCV for a patient with a given gender, age, and race/ethnicity.To do these, nationally representative data from the National Health Examination and Nutrition Survey for the period 2003-2004 (NHANES_03_04) [10] will be used.

Data source and study sample
Data from demographic [11] and body measures [12] files from NHANES_03_04 [10] were downloaded and merged to create a database for this study.A total of 7291 subjects were available for analysis.Detailed un-weighted sample sizes are given in Table I.The age of the subjects varied from 8 to 85 years.
Before data analysis was initiated, three smaller sub-databases were generated.The first sub-database included all those who were in the age group 8-11 years.The second sub-database included all those who were in the age group 12-17 years.The third sub-database included all adults or those who were 18 years of age or older.All three sub-databases were categorized by gender (meals, females) and race/ethnicity (non-Hispanic whites (NHW), non-Hispanic blacks (NHB), Mexican Americans (MA), and all others (OTH).The adult sub-database was further categorized into smaller age groups of 18-29 years old, 30-49 years old, 50-69 years old, and ≥ 70 years old.The adult sub-database was also categorized into standard World Health Organization BMI categories; namely, underweight (< 18.5 kg/m 2 ), healthy weight (18.5-24.9kg/m 2 ), overweight (25.0-29.9kg/m 2 ), obesity series I (30-34.9kg/m 2 ), obesity series II (35-39.9kg/m 2 ) and obesity series III (>= 40 kg/m 2 ).

Statistical analysis
Since the distributions of TBV (skewness = 0.52), RCV (skewness = 0.69), and PV (skewness = 0.36) were not normal, the data for these variables were normalized by log-transformation.SUDAAN [13] PROC DESCRIPT was used to compute geometric means (GM) and their 95% confidence intervals (95CI).Percentile points and their 95CI for TBV, RCV, and PV were calculated by method proposed by Korn and Grubbard [14] .References ranges were computed as the difference between 2.5 th and 97.5 th percentiles.GM, percentile points, and reference ranges were also computed for TBV, RCV, and PV normalized by TBW.In order to meet the requirements of a situation where exact "should be" blood volume measurements are needed for a patient with a specific gender, race/ethnicity, and age, we developed regression models using SUDAAN [13] Proc REGRESS.The regression coefficients estimated by these models can be used to compute "should be" blood volume measurements for a specific patient under care.The only objective of fitting regressions models was to provide regression coefficients for more exact computations of blood volume measurements and no attempt was made to use all independent variables that might affect blood volume measurements.In other words, maximization of R 2 was not attempted.Since, the aim of this study was to exclusively provide what "should be" blood measurements and not to test for statistical significance between various demographic groups; no tests for statistically significant differences were conducted.

Results
GMs as well as all percentiles points (PP) for TBV, RCV, and BV were substantially higher for males than for females aged 18 and over (Tables 2-3).For example, for TBV, males had a GM of 5372.5 ml and females had a GM of 3972.5 ml, a difference of about 35%.The references ranges were also much wider for males than for females.For example, for RCV, the width of reference rage for males was 1266.8 ml, and 688.5 ml for females; a difference of about 84%.While,  the differences for GMs and PPs for NHB and NHW were minimal, MA had the lowest GMs and PPs.For example (Table 2-3), for PV, 95 th PPs for NHW, NHB, and MA were 3651.8ml, 3672.1 ml, and 3467.6 ml respectively.The widths of references ranges for TBV for NHW, NHB, and MA were in the order NHB > NHW > MA.For example, the widths of references ranges for RCV for NHW, NHB, and MA were 1478.2ml, 1585.6 ml, and 1315 ml respectively.In general, particularly for TBV and RCV, GMs and PPs increased with increase in age except that for age ≥ 70 years for which TBV, RCV, and PV were almost always the lowest.The width of reference ranges was also the lowest for age ≥ 70 years.For example, for PV, the widths of the reference ranges for 18-29 years, 30-49 years, 50-69 years, and ≥ 70 years were 1866.6 ml, 1679.1 ml, 1736.5 ml, and 1075.4 ml respectively.With increase in BMI; TBV, RCV, and PV also increased.For example (Table 2), GMs for TBV for underweight, healthy weight, overweight, obesity series I, obesity series II, and obesity series III were 3720.9 ml, 4139.2 ml, 4691.4 ml, 5005.5 ml, 5177.6 ml, and 5455.6 ml respectively.
Compared to adults, for children 8-11 years old, GMs as well as almost all percentiles points (PP) for TBV and PV for males and females were similar (Table 4-5).For example, for TBV, males had a GM of 2840.3 ml and females had a GM of 2756.1 ml, a difference of only 3%.The references ranges were wider for males than for females.For example, for PV, the width of reference rage for males was 1285.1 ml, and 1148 ml for females; a difference of about 12%.For RCV, however, females had higher GMs and most percentiles than males.For example, the GM for females was 1028.3 ml, and 938.5 for males, a difference of 9.6%.However, the width for reference range for males (1210.2ml) was larger than for females (678.5 ml), a difference of about 78%.NHB, almost always, had the highest GMs and PPs as compared to NHW and MA for TBV, RCV, and PV.For example, 10 th percentile for PV for NHB, NHW, and MA was 1464.4 ml, 1447.3 ml, and 1413.3 ml respectively.The order of widths for references ranges for TBV, RCV, and PV was NHB > MA > NHW.For example (Table 4), the widths for PV for NHB, NHW, and MA were 1474.4 ml, 1465.4 ml, and 1254.1 ml respectively.
Male adolescents (12-17 Years old) had substantially higher GMs and PPs than female adolescents (Table 6-7) for TBV, CRV, and PV.For example, for TBV, 95 th PP for male adolescents was 5839.4 ml and 4404.5 ml for female adolescents, a difference of 34.8%.Also, references ranges for male adolescents were wider than for female adolescents.For example, for CRV, the width of reference range for male adolescents (1609 ml) was more than twice the width for female adolescents (611 ml).Almost always, the GMs and PPs for TBV, RCV, and CV were in the order: NHB > NHW > MA.For example, for PV, the medians or the 50 th PPs were (Table 6) 2533.8 ml, 2445.9 ml, and 2366.8 ml for NHB, NHW, and MA respectively.NHB had the widest references ranges.For example, for PV, the widths of the reference ranges were: 1772.7 ml, 1669.4 ml, 1600.3 ml for NHB, MA, and NHW respectively.The tables for TBV, RCV, and PV normalized by TBW are presented as supplementary tables S1, S2, and S3 respectively are only briefly discussed.While the TBV/TBW obtained by us (Table S1) for TBV for female adults was similar to what has been reported earlier [2] (55.2 vs. 55.1 ml/kg), TBV/TBW computed by us was substantially lower compared to reported elsewhere [2] (62.8 ml/kg vs. 73.3m/kg) for male adult.Male adults had higher GMs and PPs for TBV/TBW, RCV/TBW, and PV/TBW as compared to females.While there was not much difference for GMS between NHW and MA, NHB had the lowest GMs for TBV/TBW, RCV/TBW, and PV/TBW.Those who were 18-29 years old had the highest GMs.There was a decrease in GMs with increase in BMI for TBV/TBW, RCV/TBW, and PV/TBW.For example (Table S1), for TBV/TBW, the GMs were 73.8 ml/kg, 65.2 ml/kg, 59.6 ml/kg, 54.2 ml/kg, Tables 8 and 9 provide regression coefficients that can be used to compute "should be" log 10 of TBV, RCV, and PV.For example, in order to compute "should be" TBV for a Mexican American male aged 31 years old (age category: 30-49 years) with a BMI = 34.2kg/m 2 (BMI category: obesity series I), the computations shall be as follows: Log 10 (TBV) = 3.65440 + 0 -0.03051 + 0.11567 -0.00113 = 3.739 TBV = 10 3.739 = 5482.8ml.GM for TBV for an "unspecified" MA was 4482 ml (95CI: 4415.1 -4549.9ml) (Table 2).The computed TBV for a 31 year old male MA with a BMI of 34.2 kg/m 2 was 5482.8 ml which is substantially higher than as compared to an "unspecified" MA.This might indicate that, if possible, the regression equations provided by us may provide more accurate results.
While the GMs and PPS were not provided for other race/ethnicities in Tables 2-7 because of relatively smaller sample sizes involved, the regression coefficients in Tables 8  and 9 do provide for computations for other race/ethnicities also.

Discussion
In this paper, we have provided the missing link that is needed to make treatment decisions, if desired, based on blood volume measurements.While, the actual blood volume measurements can be estimated using existing techniques/instruments like BVA-100 [8] , we have provided what "should be" the blood volume measurements for comparison purposes.It is up to the individual physicians/researchers to decide whether "should be" GM, reference range, or some other PP is used for comparison purposes.It probably will vary from one treatment situation to another.However, in general, if a patient's actual TBV is within 95CI of GM or median (50 th percentile), the patient may be considered to have "normal" TBV.However, the use of prediction equations provided by us may provide even a better "should be" index.It should also be noted that we have provided GMs, reference ranges, and PPs for the general U.S. which does include those who may have their blood volume measurements affected by related diseases/disorders.But, these individuals are likely to be located towards the extremes of confidence intervals presented here.