Hematological indices in an older population sample: derivation of healthy reference values

Departments of
¹ Public Health and Community Medicine,
² Ophthalmology,
³ Medicine, and
⁴ Pathology, University of Sydney, Sydney, NSW 2006, Australia.
⁵ National Centre for Epidemiology and Population Health, Australian National University, GPO 4, Canberra, ACT 2601, Australia.

⁶ Institute of Clinical Pathology and Medical Research, Westmead Hospital, Westmead, NSW 2145, Australia.
^a Address correspondence to this author at: Public Health and Community Medicine, Westmead Hospital, Westmead, NSW 2145, Australia. Fax 61 2 689 1049; e-mail rossl{at}pub.health.su.oz.au.

	Abstract

Top Abstract Introduction Materials and Methods Results Discussion References

 
Factors affecting hematological values were explored, and healthy reference values were estimated from a cross-sectional survey of a population (n = 4433), ages 49 years or more, residing permanently in a defined geographic region. Nursing home residents were excluded. Details of medication use and medical history were obtained by interview, and participants were asked to return after an overnight fast for blood sampling. The participation rate was 82.4%, of whom 88.4% provided a fasting blood sample. Hemoglobin, hematocrit, and erythrocyte counts were higher in men, whereas platelet counts were higher in women. Statistical associations between each hematological index and smoking, alcohol intake, use of certain drugs, chronic disease, and high creatinine values were tested by unpaired t-tests. Separate reference groups were defined for each hematological index by excluding subjects with any of the factors found to be of importance. The resulting reference values are particularly appropriate for evaluating hematological test results in older individuals.

	Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

 
Laboratory-based investigations are subject to substantial variability arising from several sources, including differences between subjects (e.g., age, sex, and genetic variation), within-subject variation (e.g., circadian change and pathological change), variations in sample collection and handling, and laboratory measurement error (1). In interpreting an individual patient's laboratory test results, the clinician usually compares the reported values with reference values. Inappropriate reference values may increase the risk of either unnecessary additional investigations or failure to detect underlying disease. In clinical practice, reference values are often printed by the testing laboratory on the same document as the results, although their origins are rarely specified.

Reference values for a given patient are usually defined in terms of the spread of results typically encountered from similar subjects who are known to be in good health. Ideally, the ranges are derived by using statistical criteria from a random sample of comparable individuals. To ensure that the values represent those encountered in health, the results from individuals in the sample who are found to have acute or chronic disease are usually excluded. Published reference values have been derived from a variety of samples, including those from healthy volunteers, subjects attending health screening (2) or a routine medical examination (3), first-time blood donors (4), preemployment testing, and subjects in retrospective studies (5). The extent to which each of these different sampling strategies produces reference values appropriate to any given patient is not always clear.

Reference values for the elderly may differ from those in younger people (1). Deriving reference values for older patients is particularly problematic, because age-related physiological changes are also known to occur (6), and the prevalence of subclinical disease increases markedly with advancing age. Several studies have reported reference values from older populations (7)(8)(9)(10)(11)(12)(13)(14)(15)(16), but most have used nonideal selection criteria.

The aims of this study were: (a) to explore factors affecting hematological values; (b) to select a reference group free from factors found to influence hematological values; and (c) to produce healthy reference values by using data from a cross-sectional survey of an older population sample.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

 
subjects
Subjects included in this study were participants in the Blue Mountains Eye Study, a cross-sectional survey of the prevalence and causes of visual impairment that has been described in detail elsewhere (17). All permanent residents (excluding nursing home residents) of two postal code areas in the Blue Mountains, west of Sydney, Australia, ages 49 or above at the start of the study (1992), were invited to participate. Participants completed a demographic and medical questionnaire and attended an eye examination. They were also asked to return within 4 weeks for blood tests after an overnight fast. The study was approved by the relevant Ethics Committee, and all participants gave informed consent.

blood collection and measurements
Fasting blood was collected in the morning, centrifuged on site, and then transported for laboratory analysis within 4 h of collection. All tests were performed by the Institute of Clinical Pathology and Medical Research at Westmead Hospital. The hematological indices analyzed were hemoglobin (Hb),¹ hematocrit (HCT), erythrocyte count (RBC), mean corpuscular volume (MCV) and mean corpuscular Hb (MCH), leukocyte count (WBC), platelet count, and creatinine concentration. Hematological values were measured on a Technician H2 hematology analyzer. Creatinine measurement was performed on an Hitachi 747 biochemistry analyzer. The laboratories were fully accredited under the Royal College of Pathologists Australasia/National Association of Testing Authorities Australasia medical registration program appropriate for Australian laboratories. With the use of standard test reagents, the laboratory intraassay CV ranged from 0.3% for RBC to 1.5% for MCH, whereas the interassay CV over a 3-week period ranged from 0.6% for MCV to 5.3% for platelet count.

definitions
Factors that may have had an association with hematological values were sought from demographic and medical questionnaire responses. Current smokers were defined as those currently smoking manufactured cigarettes, hand-rolled cigarettes, cigars, or pipe tobacco. Exsmokers were those who had ever regularly smoked cigarettes, cigars, or a pipe. Alcohol drinkers were those who consumed three or more drinks on the day they drank for 3 or more days in a week. Drug users were defined as those who regularly took tablets for gout or arthritis (nonsteroidal anti-inflammatory drugs), chloroquine, plaquenil, or oral steroids. Subjects who reported having diabetes mellitus or who reported treatment with insulin, thyroid disease treated with thyroxine, cancer, rheumatoid arthritis, or gouty arthritis were defined as having chronic disease. The threshold for defining high creatinine values was set at 150 µmol/L.

data quality checking
We identified, from the computerized data, all subjects who had any missing values or any values above the 97.5th percentile or below the 2.5th percentile for each hematological index. Their original paper records were checked, and any coding mistakes were rectified to produce the data set used for all analyses. Extreme values were not excluded.

statistical methods and presentation
All statistical analysis was performed by using the Statistical Analysis System package (18). For each hematological index, potential confounding factors of interest were smoking status, high alcohol intake, medication, self-reported chronic disease, and high creatinine values as defined above. Where there were significant differences (P <0.05) between mean values by unpaired t-test, the factor was defined as being associated with the hematological index, and any subject with that particular factor present was excluded from the reference group for that particular hematological index. Any given subject may have been excluded from the reference group for some indices but not for others. In this way, the reference sample sizes for each hematological index were maximized, because very few participants, particularly at advanced ages, would have been free from all of the factors thought to be potentially associated with changes in hematological values.

The distribution of each hematological characteristic in the reference subjects was tested for departure from the gaussian distribution by using the Shapiro–Wilk statistic (18). Hematological values in reference subjects were expressed in two ways, both of which are in general use in the literature (19). Assuming a gaussian distribution, the range extending 2 SDs each side of the mean value (Mid-2SD) is expected to cover slightly more than 95% of all values. Ranking the observations from smallest to largest, the range of values from the 2.5th to the 97.5th percentile (Mid-95%) covered the central 95% of values precisely.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

 
Of the 4433 eligible residents identified in the two relevant areas, 3654 participated in the study, giving a participation rate of 82.4%. Of these 3654 participants, results from fasting blood tests were available from 3219 (88.4%), of whom 1837 were women (57%). The median and mean age was 66 years, and the range was from 49 to 97 years. With the exception of Hb, HCT, and RBC in men, the distribution of hematological values from reference subjects was significantly different from the gaussian distribution (P <0.0001).

sex differences
There were highly statistically significant differences (P =0.0001) for all hematological indices between women and men; therefore, all subsequent analyses were stratified by sex. Men had higher Hb, HCT, RBC, WBC, and creatinine values than women. Platelet counts were significantly higher in women than men. Sex differences were less marked for MCV and MCH. Table 1 shows descriptive statistics for each hematological index by sex from the entire sample.

effects of exclusions
A summary of statistically significant associations between each potential confounding factor and each hematological variable by unpaired t-test is shown in Table 2 . Table 3 shows these by sex from subjects who satisfied the criteria for inclusion in the reference group for each hematological index. Note that the number of reference subjects varied for each index because of the variable number of subjects excluded. For most indices, the reference subjects had smaller ranges and lower SDs compared with the sample as a whole. For example, the effect of subject exclusion on WBCs included a substantial lowering of the mean value as well as compression of the range.

comparison with other studies
Table 4 compares these reference values with those of other similar published studies. Many of these had relatively small sample sizes. Results from this study were broadly similar to those from the largest comparable report, although exclusion criteria differed (12). The Hb values, particularly in women in this study, were generally higher than other studies. This is particularly the case in comparison with studies that did not exclude subjects with chronic disease, tobacco use, or drug use. The upper limit of the range for WBC was lower, particularly for men and particularly in comparison to samples not selected by health status. Reference values for other indices were similar to those reported from other studies with large, well-selected samples.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

 
reference sample selection and methods
Choice of sampling frame and subject selection criteria are central issues in deriving reference values (20). The choice of a reference population lies between a natural population, unselected for disease, and a selected population of healthy individuals as defined by specific exclusion criteria (9). Both methods have been used in studies published previously. Those studies that use natural populations unselected for disease (8)(9)(10)(11)(12)(13) are likely to have included a greater number of truly abnormal values from unhealthy participants. These abnormal values are likely to result in wider ranges, particularly in the directions associated with disease. Increased WBC is associated with increased mortality and may serve as a marker for chronic disease (21). The higher upper reference limit values for WBC shown in Table 4 from relatively poorly screened samples (e.g., Woo et al. (16) and Arumanayagam et al. (14)) illustrate this point.

In this report, the reference sample for each hematological index was selected by using explicit, statistically based criteria that gave relatively large sample sizes and maximized the precision of the estimates while minimizing bias from variation because of life-style, medication, and disease. Sample handling and laboratory sources of variation were minimized because all specimens were handled uniformly, and all tests were performed by one pathology laboratory.

generalizability of the results
There were substantial differences between women and men for all hematological indices; therefore, separate reference values are needed. The demographic characteristics of the sample were not markedly different from the New South Wales or Australian over-50 population (22). Although the population of the area has a higher proportion of new arrivals through retirement, the effect of this retirement cohort is unlikely to be large because there are no differences in socio-demographic, disease, risk factor, and health service utilization profiles of new arrivals within 2 years compared with those of the same-age-group residents for a longer period (23).

Most of the participants live at altitudes between 700 and 1000 m above sea level. Higher altitudes are known to increase the RBC, without changes, or with a slight decrease in Hb and HCT, and with substantial decrease in MCV (24). However, compared with other published data from older populations, there was no evidence of a substantial altitude effect on HCT or MCV in this sample (Table 4 ).

Systematic differences between studies reporting reference ranges may arise as a result of the use of different laboratory techniques and differences in sample age-distribution patterns. Many previous studies have reported only parametric ranges. In our data, the majority of the hematological indices were not gaussian distributed in the reference subjects; therefore, the use of nonparametric (e.g., Mid-95%) ranges is preferred.

The hematological reference ranges presented here are derived from a large, representative population sample with a high participation rate; therefore, the results are likely to be generalizable to Caucasian subjects in other developed Western countries where similar laboratory methods are in use.

In conclusion, given that the values reported here were derived from a large population sample and that subjects were excluded by using rational, statistically driven criteria, we argue that it is entirely appropriate to compare hematological test results from older adults with the reference values derived from this study. Data from healthy young adults may not be ideal for the evaluation of test results from an older patient if reference values from healthy older subjects are available. Data derived from unselected populations, or populations selected for disease by virtue of having laboratory investigations performed, are likely to yield biased values.

	Acknowledgments

 
Funding was provided by the National Health and Medical Research Council of Australia. Support for this project from the Institute of Clinical Pathology and Medical Research of the Western Sydney Area Health Service is gratefully acknowledged.

	Footnotes

 
¹ Nonstandard abbreviations: Hb, hemoglobin; HCT, hematocrit; RBC, erythrocyte count; MCV, mean corpuscular volume; MCH, mean corpuscular hemoglobin; WBC, leukocyte count; Mid-2SD, range extending 2 SD above and below the mean; and Mid-95%, range extending from 2.5th percentile to 97.5th percentile values.

	References

Top Abstract Introduction Materials and Methods Results Discussion References

 

1. Soldberg HE. Approved recommendation (1986) on the theory of reference values, part 1. The concept of reference values. Clin Chim Acta 1987;165:111-118. [ISI][Medline] [Order article via Infotrieve]
2. Bain B. Normal ranges. Blood cells: a practical guide 2nd ed. 1995:148-152 Blackwell Science Ltd. London. .
3. Lipschitz D, Mitchell C, Thompson C. The anemia of senescence. Am J Hematol 1981;11:51-54.
4. Fielding J, Karabus C, Brunstrom G. Storage iron depletion in male blood donors: its significance for iron status in women. J Clin Pathol 1981;21:402.[Abstract/Free Full Text]
5. Kalchthaler T, Rigor M. Anemia in institutionalized elderly patients. J Am Geriatr Soc 1981;28:108-113.
6. Faulkner WR eds. Geriatric clinical chemistry: reference values 1994 AACC Press Washington, DC. .
7. Caird F. Problems of interpretation of laboratory findings in old age. Br Med J 1973;iv:348–51..
8. Silagy C, Salem H, Campion K, McNeil J, Robert J, Donnan G, et al. Haematological profile of healthy elderly Australians. Med J Aust 1992;157:96-100. [ISI][Medline] [Order article via Infotrieve]
9. Davis R, Kelsall G, Stenhouse N, Woodliff H. "Normal" haematological values in Western Australia. Med J Aust 1971;:1015-1017. [ISI][Medline] [Order article via Infotrieve]
10. Kelly A, Munan L. Haematologic profile of natural populations: red cell parameters. Br J Haematol 1977;35:153-159. [ISI][Medline] [Order article via Infotrieve]
11. Htoo M, Kofkoff R, Freedman M. Erythrocyte parameters in the elderly: an argument against new geriatric normal values. J Am Geriatr Soc 1979;27:547-551. [ISI][Medline] [Order article via Infotrieve]
12. Jernigan J, Gudat J, Blake J, Bowen L, Lezotte D. Reference values for blood findings in relatively fit elderly persons. J Am Geriatr Soc 1980;28:308-314.
13. Yip R, Johnson C, Dallman P. Age-related changes in laboratory values used in the diagnosis of anemia and iron deficiency. Am J Clin Nutr 1984;39:427-436. [Abstract/Free Full Text]
14. Mattila K, Kuusela V, Pelliniemi T-T, Rajamaki A, Kaihola H-L, Juva K. Haematological laboratory finding in the elderly: influence of age and sex. Scand J Clin Invest 1986;46:411-415. [ISI][Medline] [Order article via Infotrieve]
15. Arumanayagam M, Lam Y, Swaminathan R, Donnan S, Horn B. Blood cell values in healthy Hong Kong Chinese adults. Clin Lab Haematol 1987;9:263-269. [ISI][Medline] [Order article via Infotrieve]
16. Zauber N, Zauber A. Hematologic data of healthy very old people. JAMA 1987;257:2181-2184. [Abstract]
17. Woo J, Arumanayagam M, Ho S, Swaminathan R. Haematological indices and the prevalence of anaemia in an elderly Chinese population. Pathology 1989;21:31-34. [ISI][Medline] [Order article via Infotrieve]
18. Mitchell P, Smith W, Attebo K, Wang JJ. Prevalence of age-related maculopathy in Australia: the Blue Mountains Eye Study. Ophthalmology 1995;102:1459-1460.
19. SAS Institute. Statistical Analysis System, 6.12 ed. Cary, NC: SAS Institute, 1995..
20. Dybkær R. Approved recommendation (1986) on the theory of reference values, part 6. Presentation of observed values related to reference values. Clin Chim Acta 1987;170:S33-S42.
21. Weiss S, Segal M, Sparrow D, Wager C. Relation of FEV₁ and peripheral blood leukocyte count to total mortality: the Normative Aging Study. Am J Epidemiol 1995;142:493-498. [Abstract/Free Full Text]
22. . Australian Bureau of Statistics. 1991 census of population and housing 1992 Australian Government Publishing Service Canberra. .
23. Smith W, Mitchell P, Leeder S. Selection bias from sampling frames: telephone directory and electoral roll compared with door-to-door population census. Aust J Public Health 1997;21:127-133.
24. Ruiz-Arguelles G, Sanchez-Medal L, Loria A, Piedras J, Cordova M. Red cell indices in normal adults residing at altitudes from sea level to 2670 meters. Am J Hematol 1980;8:265-271. [ISI][Medline] [Order article via Infotrieve]