# Nosological level of evaluation of laboratory results

The clinician should know, understand and take into account the influence of the conditions for sampling, storage, transportation of biomaterial samples, as well as biological, analytical and iatrogenic variations on the results of laboratory studies. On the other hand, its most important duty is to take into account the influence of pathological factors that determine the deviation of laboratory results beyond the limits of "normal values" or reference intervals, that is, the actual analysis of the pathological variation at the nosological level of the evaluation of the laboratory result. In order to draw conclusions from the pathological results of laboratory tests at the nosological level, the clinician needs additional information about the characteristics of these tests in patients of different groups. In particular, data on the degree of pathognomonic changes in the magnitude of the laboratory indicator for a particular pathology, the sensitivity, specificity and prognostic value of the result of a laboratory test are needed. In addition, it is necessary to know the critical values of the results of laboratory tests, in which immediate medical attention is required.

The nosological level of the evaluation of the results of laboratory studies implies the presence of a connection of the revealed abnormalities in the analyzes with a certain pathology.

The degree of pathognomony of laboratory deviations is very variable, since the forms and severity of the pathological process itself are essentially different from one case of the disease to another. Some laboratory tests, closely related to a specific function of the organ, tissue, organism, disturbed pathological process, are almost selective.

Detection of increased activity in the pancreatic a-ami-lase in the blood indicates a pancreatic injury, since this isoenzyme can be synthesized only in it. The frequency of detecting elevated concentrations of T and I troponins in blood with myocardial infarction( MI) is very high, since these proteins play a crucial role in the functioning of the contractile system of the heart muscle. Pato-gnomonicity of deviations in the results of laboratory tests is very indicative for genetically determined metabolic disorders( phenylketonuria, galactosemia, etc.).

However, the process of establishing a diagnosis is imperfect: as a result, the clinician can only assume that the diagnosis is correct, rather than state it with certainty. Previously, clinicians expressed a degree of confidence in the clinical diagnosis, anticipating its wording with the words "excluded. .." or "possibly. ..".Now, increasingly, this confidence in the diagnosis is expressed in terms of probabilities. Therefore, the doctor should understand the statistical significance of the diagnostic value of laboratory tests in various situations. As a rule, it helps to reduce the degree of uncertainty of the diagnosis with the help of a particular result of a laboratory test, in a number of cases, to be convinced of its uncertainty, and sometimes - only to realize the degree of its uncertainty.

The relationship between the result of a laboratory test and an accurate diagnosis is schematically shown in Fig. The test result can be either positive( pathology) or negative( normal), and the disease can either be or be absent. There are four possible interpretations of the test results - two true and two false. The correct answer is a positive result in the presence of a disease or a negative result in its absence. On the contrary, the answer is erroneous if the result of the test is positive( false positive), although the person is healthy, or negative( false-negative), although the person is sick.

The main characteristics of the laboratory test are its diagnostic sensitivity and specificity. The probability of a positive result of a diagnostic test in the presence of a disease is called the sensitivity of the method, and the probability of a negative result in the absence of a disease is its specificity. A sensitive test rarely "misses" patients who have a disease. A specific test, as a rule, "does not relate" healthy people to the category of patients. Practically these characteristics of laboratory tests are determined on the basis of statistical analysis of the arrays of clinical and laboratory research results and mathematically characterize the integral effect of the pathognomonic character of the laboratory indicator for a certain form of pathology. At the base of the

Fig. The relationship between the results of the laboratory test and the presence of the disease

Fig. The relationship between the results of the laboratory test and the presence of the disease

takes the distribution of research results in accordance with the data given in the table. In most cases, these characteristics coincide, being truly positive( the disease exists and the test confirms it) or true negative( there is no disease and the test excludes it).However, the results can also be false-negative( the disease is there, but the test excludes it), and false positive( there is no disease, but the test confirms it).

Table Criteria for evaluating the results of laboratory studies

Table Criteria for evaluating the results of laboratory tests

For a clinician, a sensitive test is especially informative when its result is negative( that is, excludes healthy patients), and a specific test is most effective when its result is positivethere are patients among the healthy).Therefore, sensitive tests are recommended to be used in the early stages of diagnostic search to narrow its scope, when there are many possible options and diagnostic tests allow excluding some, that is, to conclude that these diseases are unlikely. Specific tests are needed to confirm the diagnosis, based on other data. The results of a highly specific test should not be positive in the absence of the disease. Such tests should be used if a false positive result can cause harm to the patient. For example, before assigning chemotherapy to a patient with a malignant neoplasm, associated with risk, emotional trauma, morphological confirmation of the diagnosis is necessary, since an increase in the concentration of tumor markers and data from other methods of investigation is insufficient.

The clinician should understand that the diagnostic sensitivity and specificity of the test depend on the value of the reference range, that is, the choice of the separation point, in which any test result value above this point is considered a pathology. Clinical goals can influence the choice of the point of separation. If we take the point "A" for the separation position, then the test will have 100% sensitivity to the disease and very low specificity. If we use the "C" for this purpose, then the test will have 100% specificity, but very low sensitivity. Therefore, for most tests, the separation point( "B") is determined by the reference range, that is, the range of test results that are in the range of + 2S with an average value of "B".In some cases, the value of the separation point varies depending on the purpose of the study, which increases either sensitivity or specificity.

Fig. Hypothetical distribution of test results among healthy and sick

Fig. Hypothetical distribution of test results among healthy and sick

The sensitivity and specificity of the study should be considered when deciding whether to prescribe this test. However, if a test is assigned and its results obtained( positive or negative), the concepts of sensitivity and specificity lose meaning. For the clinician, the most important issue now is the problem - how great is the probability that the disease is actually present, if the test result is positive, or with what reliability it is possible to exclude it if the test is negative. These questions can be answered using the PCRP and CER.

PCRP - the probability of having a disease with a positive( pathological) test result. PCR - the probability of absence of disease with a negative( normal) test result. Knowledge of the predictive value( PC) of test results allows the doctor to answer the question: "What is the probability that this patient suffers / does not suffer from a certain disease if his test result is positive / negative?"

PC test in relation to a particular disease( post-test probability) depends not only on its specificity and sensitivity, but also on the prevalence of the disease itself. PCRR with respect to a particular disease can be calculated by the following formula.

where: ЧЗ - sensitivity of the test;РЗ - prevalence of the disease;CT - specificity of the test.

The prevalence of the disease is also called a pretest probability, that is, it is the probability of identifying the disease before the test results become known. How to assess the pretest probability of the disease in a patient in order to calculate the PC of a particular test result? There are several sources of information: medical literature, archives of medical institutions, personal experience of each doctor.

PC is associated with the reference value and depends on the ratio of the true test results( both positive and negative) and false. The more sensitive the test, the higher the HRC of its negative result( that is, the doctor's confidence increases that negative test results reject the presence of the disease).Conversely, the more specific the test, the higher the HRC of its positive result( that is, the doctor can more confidently assume that the positive test results confirm the alleged diagnosis).Since the prevalence of the disease affects the PC test, the latter inevitably depends on the conditions of its application. If the positive results of even a highly specific laboratory test are obtained in a population with a low probability of disease, they will turn out to be mostly false positive. Similarly, the negative results of a highly specific test obtained in a population with high chances of having a disease are likely to be false-negative. Thus, the interpretation of the PC positive or negative results of the laboratory test varies depending on the prevalence of the disease. A test with a high PCR is effective in examining a contingent with a high prevalence of pathology, for example, for patients in a specialized department of a hospital, whereas in an outpatient study, a test with a high

is more useful. Similarly, the degree of probability of diagnosis affects the PC test( if the probability of diagnosis is low, the value of the test with the CRR increases, if higher, the test with PCR is more valuable).

The relationship between sensitivity, specificity and PC of laboratory tests is presented in Fig.

If we imagine a population in which no one has the disease under consideration, then all the positive results in such a group, even with a very specific test, will be false positive. Therefore, when the prevalence of the disease tends to zero, the PCR of the test also tends to zero. Conversely, if this disease exists in everyone in the studied population, all the negative results of even a highly sensitive test will be false-negative. When the prevalence tends to 100%, the test PTSR tends to zero.

So, if you assign studies to search for pheochromocytoma in all patients with hypertension, the PC test with a high PCR is lower than in the case of the same study for patients with hypertension with predominantly paroxysmal course and accompanied by other characteristic manifestations of hyperkatecholamineemia. Let us illustrate the above arguments with the calculations of the PCRP in the diagnosis of pheochromocytoma for the method for determining in the urine an increased concentration of free norm-tanephrine.

Pheochromocytoma is found in approximately 0.3-0.7%( pretest probability) of patients with arterial hypertension, and among malignant current forms - in 10-15% [Dedov II, 1995].The sensitivity of the method for the determination of free normetanephrine in daily urine for the diagnosis of pheochromocytoma is 89-100%, specificity 98% [Wallach J. M.D., 1996].Initially, we will calculate the PCR for this method if it were assigned to all patients with hypertension. For the sensitivity of the test, take 90%( 0.9), for prevalence - 0.5%( 0.005).

When calculating PCR for this method, in patients with malignant current forms of hypertension for pretest probability we take 12%( 0.12).

This example shows that the pretest probability of the disease has a large impact on post-test probability( PC).From the data given below( Table) it follows that when using a test with 90% sensitivity and specificity, the post-test probability can vary from 8 to 99%, depending on the pretested probability. Moreover, as soon as the pretest probability of the disease decreases, it becomes less likely( post-test probability) that

Fig. Relation of sensitivity, specificity and PC of laboratory tests in the solution matrix [by Gornall A. G., 1980]

Fig. The relationship between sensitivity, specificity and PC of laboratory tests in the solution matrix [according to Gornall A. G., 1980]

a patient with a positive test is sick, and it is more probable that the test result is false positive.

In his studies R. Fletcher et al.(1998) showed that if a prostatic arterial hypertension( PSA) is used to diagnose prostate cancer in all elderly men who have no symptoms, and the prostate cancer incidence is 6-12%( pretest probability), then the post-test probabilitywill be only 15% at a PSA concentration of 4 ng / ml( sensitivity 90%, specificity 60%) and above. In a PSA study in a higher-risk group( with symptoms or suspicious results of digital rectal examination) with pre-

, a 26% probability of post-test probability was 40% at the same PSA concentration. Finally, in the determination of PSA in patients with a detected node in the prostate gland with a rectal examination, the presence of bone pains, the rarefaction in the bones during X-ray examination, the pretest probability was 98%, and post-test - 99%.

Table Impact of pretest probability on post-test probability of disease using 90% sensitivity test and 90% specificity of

Table Effect of pretest probability on post-test probability of disease using 90% sensitivity test and 90% specificity of

This example shows that pretest probability has a hugeinfluence on post-test and that studies give more information when the diagnosis is really uncertain( pretest probability is about 26%) than in the unlikely( pretest probability 6-12%) or almost unquestionable( pretest probability 98%) diagnosis.

The above reasoning shows that assessing pretest probability is as important an important part of the diagnosis process as the sensitivity and specificity of the laboratory test. In this regard, in clinical practice it is very important to choose the optimal method of investigation, since a test with a lower sensitivity and specificity in an experienced physician( based on personal experience it has a high pretest probability) may have the same post-test probability as a test with greater sensitivityand the specificity of a less experienced clinician.

Let's demonstrate this on the example of acute pancreatitis diagnosis. In Table. The sensitivity and specificity of the main tests used for the diagnosis of acute pancreatitis are given.

Table Sensitivity and specificity of laboratory tests for the diagnosis of acute pancreatitis

Table Sensitivity and specificity of laboratory tests for the diagnosis of acute pancreatitis

The pretest probability of a patient having acute pancreatitis( according to the clinician's conclusion taking into account an anamnesis, a clinical picture of the disease, objective examination data) can vary very widely- from 7 to 59%, averaging 21% [Buchler MW et al., 1999].This means that acute pancreatitis is present in 1 out of 5 patients with suspected disease. Taking into account this( 21%) pre-test probability of the presence of the disease( or its absence - 79%) and taking into account the sensitivity and specificity presented in Table.the post-test probability of acute pancreatitis will be 65% if it is based only on a positive result of a study of total amylase in the serum( Table).This post-test probability is not sufficient to confirm the diagnosis of acute pancreatitis. If the activity of amylase is normal, post-test probability will be only 6%.The indicators are better for pancreatic amylase and even better for lipase. If the serum lipase activity is higher than normal, the probability of acute pancreatitis is 86%, and with normal lipase activity it will be only 1.6%.

The activity of lipase in the blood remains elevated for a longer time than total amylase and pancreatic amylase. Thus, the diagnostic efficacy of lipase research in acute pancreatitis is significantly higher than any of the amylases, beginning with the second day of the disease. With a pretest probability of 50% and a positive result of the study of total amylase( sensitivity 83%), the post-test probability of acute pancreatitis will be already 87%.

Table Sensitivity, specificity, PCR and PTSR of laboratory tests for the diagnosis of acute pancreatitis with pretest probability of 21% [Buchler MW et al., 1999]

Table Sensitivity, specificity, PCR and PTSR of laboratory tests for acute pancreatitis in pretest probability 21% [Buchler MW et al., 1999]

These examples show that the pretest probability of the disease has a big impact on post-test probability. Several tests conducted in parallel provide, as a rule, a higher sensitivity, and therefore, a greater CRR for this pathology than each test separately.

Thus, the PC of the laboratory test( post-test probability) is the most adequate characteristic for the interpretation of its results. It is determined not only by the sensitivity and specificity of the test, but also by the pretest probability. Usually, in order to get a fairly reliable diagnosis, you have to use several laboratory tests in parallel or sequentially.

The use of the presented approaches to the evaluation of laboratory results significantly strengthens the methodological level of the clinical

clinical practice, helping to more accurately assess the probability of presence or absence of acute pancreatitis in the patient.

Another way to evaluate the effectiveness of a diagnostic test is to use likelihood ratios( OPs) that summarize the same information as sensitivity and specificity indicators and can be used to calculate the probability of a disease( post-test probability) based on a positive or negative test result.

The OP for a specific result of a diagnostic test is the ratio of the likelihood of this result in persons with a disease to the probability of the same result in persons without a disease. OP shows how many times higher or lower the probability of obtaining a given test result in patients than in healthy patients. If the test score is dichotomized( positive-negative), then its ability to distinguish between patients and healthy corresponds to two types: one type is associated with a positive test result, the other with a negative test result.

The OP positive( ORD) or negative( OPOR) result is calculated as follows:

where: ЧТ - sensitivity of the test;CT - specificity of the test.

The values of OP can be found in textbooks, medical journals and computer programs( Table) or calculated by the above formulas.

Table Examples of OD for some tests [Nicoll D. et al., 1997]

Table Examples of OD for some tests [Nicoll D. et al., 1997]

The simplest method for calculating post-test probability for pretest probability( prevalence)and OP - the use of the nomogram. It is necessary to place the ruler so that its edge passes through the points corresponding to the value of the pretest probability and OD, and to note the point of intersection with the post-test probability line.

Post-test probability can also be calculated using the following formula:

post-test chances = pre-test chances x OP.

To use the above formula, the probabilities should be translated into chances. Chances and probability( pretest or post test) contain the same information, but they express it in different ways.

For example, the prevalence of the disease( pretest probability) is 75%( 0.75), therefore, the pretest odds are:

For example, the prevalence of the disease( pretest probability) is 75%( 0.75), therefore, the pretest odds are:

BFurther, knowing the pretest odds and ODPR / OPOR, by multiplying them you can get post-test chances of having the disease if the test is positive / negative.

For example, the physician assumes that the patient has a probability of MI of 60%( pretest odds of 3: 2), and the activity of the MB fraction of KK( KB-MB) in the serum is increased( positive test).In Table.we find the DAC and the OPOR studies of the KK-MB-32 and 0.05, respectively. The post-test chances of having MI are: if the result is positive, 3/2 x 32 = 48/1 [post-test probability -( 48/1) /( 48/1) + 1 = 0.98 or 98%];with a negative result - 3/2 x 0.05 = 0.15 / 2 [post-test probability -( 0.15 / 2) /( 0.15 / 2) + 1 = 0.07 or 7%].

The main advantage of OP is that they help to go beyond the rough assessment of the results of the laboratory test( either the norm or the pathology) that the clinician faces when assessing the accuracy of the diagnostic test using only the concepts of sensitivity and specificity at a single point of separation. However, for most laboratory tests, this can not be achieved. In such situations, the position of the point of separation on a continuous transition between the norm and the pathology is established arbitrarily. The OP can be defined for any number of test results over the entire range of allowable values. Obviously, the presence of the disease is more likely with the extreme deviation of the test result from the norm than in the case of a result close to the norm boundary. With this approach, the clinician receives information about the degree of deviation from the norm, and not only about the fact of the presence or absence of the disease. When calculating the OP within a certain range of values of the results of the test, sensitivity is understood as the doctor's confidence in using a specific test result to identify persons with the disease, rather than to some degree of deviation from the norm. The same applies to specificity. Usually, more than 10 or more than 0.1 ODP allow the final diagnostic decision to be made. The values of OVD in the range of 5 to 10 and the ODP from 0.1 to 0.2 give a moderate basis for the diagnostic solution, and if they are 2-5 and 0.2-0.5, respectively, this does little to change the probability of the disease being presentpatient. In the case of OCD and OPOR from 0.5 to 2, the likelihood of having the disease in the patient does not practically change. Let's illustrate these arguments on the example of determining the concentration of thyroxine( T4) in the blood for the diagnosis of hypothyroidism( Table).

The values of OP for hypothyroidism are greatest at low concentrations of T4 and the smallest - at high concentrations. The lowest concentrations of T4( less than 4 μg / dl) were found only in patients with hypothyroidism, that is, they certainly confirm the diagnosis. The highest

The likelihood ratio

Fig. Nomogram for determining the post-test probability of the disease by pretest probability and OP [Nicoll D. et al., 1997]

Likelihood Ratio

Fig. A nomogram for determining the post-test probability of the disease by pretest probability and OP [Nicoll D. et al., 1997]

concentrations of T4( > 8 μg / dL) in patients with hypothyroidism are not observed at all, ie they exclude this diagnosis.

Thus, the OD value corresponds to a reasonable clinical practice when, in assessing the likelihood of a disease, a higher weight( or low) test result, rather than a borderline between the norm and the pathology, is given a higher weight. OP is particularly useful for determining the likelihood of having a disease when several diagnostic tests are used consistently.

Because laboratory tests are used in clinical practice, the sensitivity and specificity of which is below 100%, the probability of having the disease with only one test is often determined as not very high and not very low, between 10 and 90%.As a rule, after receiving such a result, the doctor can not stop the diagnostic search. In such situations, he tries to significantly increase or decrease the probability of detection of the disease( post-test probability) and continues the examination of the patient, applying additional tests.

When several tests are performed and the results of all are positive( pathological) or negative( normal), their meaning is obvious. Much more often it happens that the results of some tests are positive, and others - negative. Then their clinical evaluation becomes more complicated.

There are two ways to apply several tests: parallel( several tests at the same time, and the positive result of any of them is considered in favor of the presence of the disease) and sequential, taking into account the results of the previous test. In a sequential approach for diagnosis, the results of all tests should be positive, since in the event of a negative result, the diagnostic search terminates.

Several tests are assigned in parallel when rapid assessment of the condition is necessary, for example in hospitalized patients with emergency conditions or in outpatients who have come for a brief examination. An example of the parallel assignment of several tests at the same time can serve as an appointment for the study of myoglobin, CC, LDH: in a patient with suspected MI.

Several tests conducted in parallel provide, as a rule, a higher sensitivity, and therefore, a greater CER in this pathology than each test separately. At the same time, the specificity and PCR of the test are reduced. Thus, the probability that the disease will be missed decreases, but the likelihood of false positive diagnoses increases.

The parallel use of several tests is particularly useful in situations where a very sensitive test is needed, but only a few relatively low sensitivity are actually available. Due to the parallel use of several tests, the overall sensitivity is increased. The fee for such a sensitivity increase is the examination or treatment of a number of patients in whom the disease is not being studied.

Consistent use of several diagnostic tests is preferable in clinical situations in which rapid assessment of the patient's condition is not necessary, for example in outpatient practice. In addition, the consistent application of diagnostic tests is advisable if the question arises of an expensive or risky study( for example, invasive).Such a method of investigation is usually prescribed only after the positive results of the use of noninvasive methods. For example, at a high risk of giving birth to a child with Down's syndrome, a mother's blood test for a-feto protein( AFP), chorionic gonadotropin( CG), free estriol, inhibin A is first performed, which increases the probability of diagnosing fetal syndrome before

.in the serum of patients with and without hypothyroidism [Fletcher R. et al., 1998]

Table Distribution of serum T4 concentrations in patients with and without hypothyroidism [Fletcher R. et al., 1998]

76%, and only thenpregnant women are offered amniocentesis [Wald N.J. et al., 1997].Sequential application of tests in comparison with parallel reduces the volume of laboratory studies, since each subsequent test takes into account the results of the previous one. At the same time, consistent testing requires more time, since the next study is appointed only after receiving the results of the previous one.

In a series of tests, the specificity and PCRR( post-test probability) is increased, but sensitivity and PCR are reduced. As a result, the confidence of the clinician increases that a positive test result confirms the presence of a suspected disease, but at the same time the risk of missing the disease increases. The consistent application of tests is particularly useful when none of the available diagnostic methods are highly specific. If the doctor is going to apply the two tests sequentially, then it is more effective to first assign a test with more specificity.

In case of consecutive application of tests( A, B, C), OPs allow calculating post-test probability of the disease using the results of all tests: post-test chances = pretest odds x OD test A x. OP test B x OP test C.

Thus, the PC of the laboratory test( post-test probability) is the most adequate characteristic for interpreting its results. It is determined not only by the sensitivity and specificity of the test, but also by the prevalence of the disease in the population. Usually, to establish a sufficiently reliable diagnosis, you must use several laboratory tests in parallel or sequentially.