Individual Reference Intervals of Hematological and Serum Biochemical Parameters in Cynomolgus Monkeys

Abstract

Cynomolgus monkeys, one of a number of primates phylogenetically close to humans, are commonly used in animal studies. The purpose of this study was to assess biological variations in hematological and serum biochemical parameters in cynomolgus monkeys. Summary statistics and reference intervals were calculated using data from 95 male and 95 female Chinese-bred cynomolgus monkeys aged 3 to 7 years showing no abnormalities during the breeding period. Within- and between-animal variations were estimated using a random-effect analysis of variance (ANOVA), then, a simple method that applies prior information was proposed to estimate individual reference intervals. Parameters including MCV, MCH, PT, ALP, total cholesterol, and creatinine appeared to show a large between-animal variation; thus, it is considered that individual reference intervals for these parameters would be relatively small in comparison with overall reference intervals.

Reference intervals, sometimes called reference ranges or normal ranges, are important in clinical and preclinical studies to evaluate health condition indices and improve the accuracy of studies to evaluate drug safety. Extensive information on drug safety evaluation in humans is available due to the large number of studies performed (NCCLS 2000); however, no information is available on reference intervals in hematological and serum biochemical parameters in cynomolgus monkeys (Macaca fascicularis), one of a number of primates phylogenetically close to humans. The cynomolgus monkey has been adopted as a study animal more frequently in recent years to evaluate drug safety or drug pharmacology, therefore fundamental information on hematological and serum biochemical parameters in cynomolgus monkeys must be determined as soon as possible. It has been reported that, with humans, these variables vary with physiological condition (Demacker et al. 1982; Evans and Laker 1995; Widjaja et al. 1999), and it is necessary to verify whether similar variation occurs in cynomolgus monkeys.

First, summary statistics and reference intervals of hematological and serum biochemical parameters in cynomolgus monkeys were calculated to grasp distributions of these parameters. Next, within-animal variation and its ratio to total variation were estimated for each parameter to assess physiological variations in individual animals. Finally, a simple method that applies prior information was proposed to estimate individual reference intervals. This method is useful for evaluating individual health condition index and drug safety in individual animals at evaluation, on the basis of within-animal variation.

MATERIALS AND METHODS AND RESULTS

Study System

Animals

Ninety-five male and 95 female Chinese-bred cynomolgus monkeys aged 3 to 7 years were allocated. The animals were acclimated for approximately 1 week. During this period, clinical signs were observed at least once daily, and hematological and serum biochemical parameters were measured once. No animal showed abnormalities during the acclimation period, thus data from these animals obtained from 2003 to 2004 were used. Some parameters were not measured in certain animals (see Tables 2 and 3).

Animal Maintenance Conditions

Room controls were set to maintain temperature and humidity in the ranges of 23°C to 29°C and 35% to 75%, respectively, with air changes 15 times/h and a 12-h artificial light cycle. The animals were provided approximately 108 g (12 g × 9 biscuits) of solid food at approximately the same time daily, and any remaining food was removed on the following day. Drinking water, certified to meet the Water Quality Standard required by the Japanese Waterworks Law, was available ad libitum from an automatic supply.

Hematological and Biochemical Parameters

Parameters, abbreviations, units, and analysis methods are shown in Table 1. For hematological analysis, whole blood was drawn from the femoral vein with an anticoagulant. Coagulation tests were performed with 3.8% sodium citrate-treated plasma. PT and APTT were measured using coagulation analyzer CA-5000 (Sysmex, Kobe, Japan) (Kurata and Horii 2004). Other hematology parameters were determined or counted with an automated blood cell counter ADVIA 120 (Bayer Medical, Tokyo, Japan) (Moritz et al. 2004). For serum biochemical analysis, blood was drawn from the femoral vein, and serum was obtained by centrifugation after stabilization at room temperature for 20 to 60 min. Blood biochemical parameters were measured using automated analyzer BM-8 (JEOL, Tokyo, Japan). All hematological and blood biochemical analyses were conducted four times a month. Histograms of male cynomolgus monkeys are shown in Figures 1 and 2.

Histograms and nonparametric density lines show that many parameters including WBC, leukocyte fractions, AST, ALT, ALP, LDH, CPK, T.Bil., TGL, and GLU had a right-heavy tail distribution, and it appears that it would be difficult to assume that these parameters follow normal (Gaussian) distribution. Histograms for female cynomolgus monkeys are not included, but their readings were similar to those of males. Each parameter in males and females was considered to follow a similar distribution.

Reference Intervals

Because data were obtained from a relatively large number of animals, classical but simple methods were used for estimating reference intervals (Kaputo 1972; Amador 1974). When the data were considered to have followed a normal distribution based on a density plot and histogram, the interval was calculated as mean ±1.96 SD. When the data were considered to have followed a log-normal distribution, the interval was calculated as mean ±1.96 SD for log-transformed data and the estimates were then back-transformed to the original scale. Nonparametric percentile estimates (2.5%, 97.5%) (Massod 1977) were also calculated for comparison. The results for males and females are shown separately in Tables 2 and 3.

All reference intervals and nonparametric reference intervals were considered to be substantially consistent. From these statistics and histograms, a normal/log-normal distribution could be assumed, to which many common statistical analysis methods can be applied.

Within- and Between-Animal Variations

Reference intervals for the analysis parameters shown in Tables 2 and 3 are overall reference intervals obtained from all animals judged to be healthy; however, it is thought that the range of biological variation in individual animals is much narrower than that of overall data for some parameters.

To assess within- and between-animal variations, individual within-animal variance was assumed to be similar, and analysis of variance (ANOVA) was applied using animals as a random effect. $$y_{ij}=\mu +x_i+{\varepsilon }_{ij} i=1,\dots ,n; j=1,\dots m$$ y _{i j} represents a response variable for the jth observation of animal i. x _i are mutually independent random effects on animals, and ɛ _{i j} are mutually independent random errors. x _i and ɛ _{i j} are normal with mean 0, and variance ${\sigma }_0^2$ and σ², respectively. Strictly speaking, σ² is the sum of measurement error and within-animal variance; however, it can be assumed that measurement error is insignificant when compared with within-animal variance, and that σ² is approximately the same as within-animal variance.

β is the ratio of between-animal variance to variance if defined as follows: $$\beta ={\sigma }_0^2/({\sigma }^2+{\sigma }_0^2)$$ Estimated σ² and β values are also shown in Tables 2 and 3. Large β values imply that the ratio of between-animal variance in platelets, MCV, MCH, PT, ALP, T.Cho., and CRNN in both males and females is large, and that the within-animal variations in these parameters would be much narrower in comparison with overall data variations.

Individual Reference Intervals

It has become increasingly common to use a minimum number of primates, including cynomolgus monkeys, in studies to assess the effects of a drug or endocrine-disrupting chemical, and it is now common to assess effects on individual animals because of the small number used. Accordingly, overall reference intervals for parameters with large β values would be too wide to be applicable and are, therefore, impractical.

Using baseline values, which are commonly obtained from untreated healthy animals in studies, the following method was proposed to calculate individual reference intervals to allow for evaluation of the effects of a drug or endocrine-disrupting chemical.

Assuming a random sample y _i of size n from a normal distribution N(θ _i , σ²) where θ _i is from a normal distribution N(μ₀, ${\sigma }_0^2$ ). The posterior distribution of θ _i given y _i is as follows: $${\theta }_i|y_i~\text{N}({\mu }_i,\hspace{0.17em}{\sigma }_i^2)$$ where μ_i = (nŷ ${\sigma }_0^2$ + μ ₀ σ ²)/(n ${\sigma }_0^2$ + σ ²), ${\sigma }_{\mathrm{i}}^2$ = ( ${\sigma }_0^2$ σ ²)/(n ${\sigma }_0^2$ + σ ²), and ŷ = sample mean

The posterior distribution of unobserved y _new is: $$y_{\text{new}}|y_i~\text{N(}{\mu }_i,{\sigma }_{\text{i}}^2+{\sigma }^2$$ The 95% predictive limit is μ_i ±1.96 √ ( ${\sigma }_i^2+{\sigma }^2$ ) (Spiegelhalter et al. 2004). If prior information for each animal is available, an individual reference interval using this predictive limit can be constructed.

Example

Table 4 contains data on MCH, Ht, and CPK values in four male cynomolgus monkeys in the placebo group in a toxicity study aimed at investigating effects of the drug. Data from two points before dosing and five points after dosing were obtained. Using baseline values obtained before dosing, the individual reference interval (95% predictive limit) of each animal was estimated and is shown in Figure 3 along with the overall reference interval. Individual reference intervals are shown as –dotted lines and overall reference intervals are shown as dashed lines.

As shown in Figure 3, most data after dosing were within the overall reference interval. With regard to MCH, which has a large β value (= 0.8666), individual reference intervals of all animals are much narrower than the overall reference interval, and could be a more practical index to assess individual biological variation than overall reference interval. For example, data on day 14 from animal 3 was within the overall reference interval, but not within the individual reference interval of this animal. This value is suspected to be abnormal.

The individual reference interval of CPK (β = 0.2041) might be wider than the overall reference because of the small β value. As seen in the figure, this parameter contains a large within-animal variation, as do the estimated individual reference intervals.

DISCUSSION

Summary statistics and histograms reveal the distribution of hematological and serum biochemical parameters in cynomolgus monkeys. Many parameters did not follow normal distribution, and this should be taken into account when performing statistical tests such as the t test and analysis of variance (ANOVA), which assume that data follow normal distribution.

It is essential to use accurate reference intervals to assess health condition, biological variation, or effects of a drug, or endocrine-disrupting chemical on the study system. For this purpose, reference intervals for hematological and serum biochemical parameters in cynomolgus monkeys are included. Within- and between-animal variations were also evaluated, and it was found that some parameters showed large between-animal variations. As shown in the example, not only overall reference intervals but individual reference intervals calculated from only two baseline values can yield useful information for evaluation of biological variation for these parameters. More baseline information might be necessary; however, individual reference intervals of these parameters for each animal estimated by the proposed method would make it possible to assess biological variations or effects of a drug or endocrine-disrupting chemical more accurately.

It has been reported that some parameters in human beings change with age (Wright and Royston 1997; Royston 1991; Tango 1998). In cynomolgus monkeys, parameters such as ALP, RBC, and body weight are also considered to vary with age (Yoshida et al. 1986). Thus, data continuously obtained from the same animals over a long period are necessary to estimate appropriate age-related reference intervals for these parameters.