26.12.2014 Views

Clinical Biochemistry of Domestic Animals (Sixth Edition) - UMK ...

Clinical Biochemistry of Domestic Animals (Sixth Edition) - UMK ...

Clinical Biochemistry of Domestic Animals (Sixth Edition) - UMK ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

8<br />

Chapter | 1 Concepts <strong>of</strong> Normality in <strong>Clinical</strong> <strong>Biochemistry</strong><br />

patients on the basis <strong>of</strong> analyte values, the error rates, are<br />

shown, respectively, as vertically and horizontally shaded<br />

areas in Figure 1-5 . The sensitivity <strong>of</strong> the diagnostic or<br />

decision process using reference values is the probability<br />

<strong>of</strong> deciding that a truly diseased animal is diseased on the<br />

basis <strong>of</strong> the given reference value and is equal to 1 minus<br />

the vertically shaded area <strong>of</strong> Figure 1-5 . The specificity<br />

<strong>of</strong> the decision process is the probability <strong>of</strong> deciding that<br />

a truly normal animal is normal and is equal to 1 minus<br />

the horizontally shaded area <strong>of</strong> Figure 1-5 . It is possible to<br />

change the reference values to increase the sensitivity <strong>of</strong><br />

the test, but such an action will also result in a reduction in<br />

the specificity <strong>of</strong> the test.<br />

Example 2<br />

Type III diabetic dogs have the chemical form <strong>of</strong> diabetes mellitus<br />

generally regarded as the first level <strong>of</strong> development <strong>of</strong><br />

the disease <strong>of</strong>fering the highest likelihood “ for successful oral<br />

hypoglycemic therapy or dietary therapy ” (Kaneko, 1977 ) . Thus,<br />

it would be useful to distinguish type III diabetic dogs from<br />

normal dogs. Using the sample mean [155.6mg/dl (8.63mmol/<br />

liter)] and standard deviation [32.0mg/dl (1.77mmol/liter)]<br />

<strong>of</strong> the plasma glucose values given by Kaneko (1977) for five<br />

dogs with type III diabetes mellitus as reasonable estimates <strong>of</strong><br />

the corresponding parameters for the population <strong>of</strong> dogs with<br />

type III diabetes mellitus, and assuming that this population<br />

distribution is approximately Gaussian, a comparison <strong>of</strong> this<br />

distribution <strong>of</strong> glucose values can be made with that for the<br />

population <strong>of</strong> normal dogs described by the approximately<br />

Gaussian distribution with parameter estimates given in<br />

Figure 1-2 [ μ x 96.4mg/dl (5.35mmol/liter) and σ x 14.6mg/dl<br />

(0.81 mmol/liter)]. These two distributions are those shown in<br />

Figure 1-5 ; they have reasonably good separation with moderate<br />

overlap. Based on this information, a diagnostic procedure<br />

is proposed whereby a dog entering the clinic with a glucose<br />

value above 125.1mg/dl (6.94mmol/liter), the upper limit <strong>of</strong><br />

the normal reference interval, will be flagged as possibly having<br />

type III diabetes mellitus thereby indicating need for more<br />

follow-up. (Note: This is an oversimplification <strong>of</strong> actual practice<br />

because a diagnostic decision <strong>of</strong> this type would be based<br />

on additional information, such as the animal’s glucose tolerance<br />

and insulin response, making the decision rule and subsequent<br />

error calculations more complex.) This is an example<br />

<strong>of</strong> a one-sided diagnostic procedure because a dog with a<br />

glucose value below the lower limit <strong>of</strong> the reference interval<br />

would not be considered as having type III diabetes mellitus.<br />

If a dog actually having type III diabetes mellitus has a glucose<br />

value below the upper limit <strong>of</strong> the reference interval, the<br />

diagnostic procedure will make a mistake in deciding that the<br />

dog is normal. The probability <strong>of</strong> making this mistake is 0.170<br />

or 17.0%, the area to the left <strong>of</strong> a glucose value <strong>of</strong> 125.1 mg/<br />

dl in the distribution <strong>of</strong> glucose values for dogs having type<br />

III diabetes mellitus or the area to the left <strong>of</strong> the corresponding<br />

z-value, z (125.1 155.6)/32.0 0.953, for the standard<br />

Gaussian distribution (see Section II.B).<br />

[This probability can be found by interpolating from<br />

Table D in Daniel (2005) or from MINITAB Release 14.13<br />

using the reverse <strong>of</strong> the procedure described above for<br />

generating Table 1-2 . The z -value 0.953 is placed in a<br />

column <strong>of</strong> a MINITAB worksheet and the following commands<br />

given:<br />

Calc (from the main menu <strong>of</strong> MINITAB) → Probability<br />

Distributions → Normal Distribution . Within the Normal<br />

Distribution dialog box , Cumulative probability is<br />

selected, Mean is set to 0.0, Standard deviation is set to<br />

1.0, and the column <strong>of</strong> the worksheet containing the z -value<br />

is selected and placed in the Input column : Hit OK .]<br />

The clinician may be interested in determining the<br />

sensitivity and the specificity <strong>of</strong> the diagnostic procedure.<br />

The sensitivity is 1 0.170 0.830 or 83.0%. A dog that<br />

actually is normal but has a glucose value greater than<br />

125.1 mg/dl would be incorrectly classified by the proposed<br />

diagnostic procedure as having type III diabetes mellitus.<br />

The probability <strong>of</strong> making this type <strong>of</strong> error is 0.025 or<br />

2.5%, which is the area to the right <strong>of</strong> a glucose value <strong>of</strong><br />

125.1 mg/dl in the distribution <strong>of</strong> glucose values for normal<br />

dogs or the area to the right <strong>of</strong> the corresponding z-value,<br />

z (125.1 96.4)/14.6 1.96, for the standard Gaussian<br />

distribution (from Table 1-2 or using MINITAB as shown<br />

earlier). The specificity <strong>of</strong> the diagnostic procedure<br />

is 1 0.025 0.975 or 97.5%.<br />

F . Predictive Value <strong>of</strong> a Decision Based on<br />

a Reference Interval<br />

A useful quantity is the probability that a patient having a<br />

reference value outside the normal interval actually has the<br />

disease. This is known as the predictive value <strong>of</strong> a positive<br />

diagnosis , Prob(D | ). Interest could also be in determining<br />

the probability that a patient having a reference value<br />

within the normal interval is actually nondiseased or the<br />

predictive value <strong>of</strong> a negative diagnosis , Prob(D | ). The<br />

predictive value depends on the sensitivity, specificity, and<br />

prevalence ( p ) <strong>of</strong> the disease as is shown in the following<br />

equations:<br />

Prob(D )<br />

p sensitivity<br />

<br />

psensitivity ( 1 p)<br />

( 1 specificity)<br />

Prob(D )<br />

( 1<br />

p)<br />

specificity<br />

<br />

( 1 p) specificity<br />

p(<br />

1sensitivity)<br />

Figure 1-6 demonstrates the extent to which the predictive<br />

value <strong>of</strong> a positive diagnosis changes with the prevalence.<br />

In general, larger changes are seen in the predictive<br />

value <strong>of</strong> a positive diagnosis for smaller changes in the

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!