**Introduction**

**Table 1. Preconditions for the formulation of a reference interval in healthy subjects.****IR establishment**

_{min}(where SD

_{min}is the smallest DS of the subset of the DS), then the partition is recommended.

_{min}, then the partition is not recommended.

_{min}< difference < 0.75 × SD

_{min}), the arguments should differ from the purely statistical ones, as this could be due to genetic differences, i.e. to situations which are not routinely assessed.

**Selection of the reference population**

**Table 2. Exclusion criteria for the formulation of reference range in the general population.**

**Table 3. Criteria for the creation of subgroups of reference subjects.**

**Table 4. Reference intervals: Criteria for distributions in the different age group.**

**Statistics**

**Figure 1. The normal or Gaussian distribution.***goodness-of-fit test*tests. Among them, the most known and used is the Kolmogorov-Smirnov, although its real discriminant power is questioned by some researchers, especially when the parameters of the distribution are estimated based on data rather than being specified

*a priori*. Afterwards, other tests have been proposed that are best suited for this purpose, among them the test of Shapiro-Wilks (for distribution of samples greater than 2,000 subjects it should be replaced by the test for normality of Stephen) and the test of D’Agostino-Pearson. None of these tests can however indicate the type of non-normalityobserved in the case where the distribution is showing tendency to asymmetry (skewness) and kurtosis or both (Figure 3).

**Figure 2. Non-Gaussian distribution (non-normal).**

**Figure 3. Tendency to asymmetry (skewness).**

**Figure 4. General forms of kurtosis.**

**Figure 5. Interval of reference: Gaussian curve.**^{th}and the 97.5

^{ th}percentile (Figure 6). Even in these cases the values below and above these limits are considered “out of normality”. A widely diffused but not supported by solid bases opinion is that the reference interval from Gaussian and non-Gaussian distributions represents the values of individuals to be referred to (i.e., “the normal individuals”) and that the areas at the “tails” of the curve represent individuals whose values are to be rejected as “out of normality.” This is a misconception, because (18):

**Figure 6. Interpercentile intervals: nonparametric distribution.****Longitudinal comparison of laboratory results**

_{p}× 2

^{1/2 }× (CV

_{a}+ CV

_{w})

^{1/2};

*z*is the probability density function (generally 1.96 at P = 0.05),

_{p}*CV*is the intra-individual biological variability and

_{w}*CV*is the variability of analytical testing. RCV shows some special additional benefits to get information on the status of patients, particularly in the monitoring of clinically stable and well controlled conditions, such as the prognosis of the crisis of rejection in kidney transplant patients, monitoring of oral anticoagulant therapy (OAT), the glycated hemoglobin (A1c) in diabetes and other conditions (20–23). RCV is only applicable when CV

_{a}_{a}< 0.5 < CV

_{w}. Close monitoring of analytical quality is needed for, especially when the time between the first test and the next is rather long such as for glycated hemoglobin.