1. e. Choice A represents systematic bias. Choices B and C represent both systematic bias and a trend. Choice D represents a shift and demonstrates excellent precision. Choice E represents a system demonstrating poor precision.

2. a. Another way of asking the same question is: “What is the probability that three consecutive values will fall greater than 2 SD above the mean or greater than 90 mg/dL?” This question can be viewed as a normal (Gaussian) distribution with a 95% confidence interval. The probability that one value will fall outside of this 95% confidence interval is 5%. However, the probability of one value falling greater than 2 SD above the mean is 2.5% (1/40). Therefore, the probability that three consecutive values will all fall greater than 2 SD above the mean is the product of each of the individual probabilities or (1/40)(1/40)(1/40) or 1 in 64,000. This event is not likely to occur by chance alone.

3. b. Chi-squared analysis evaluates observed and expected observations to determine if populations are related. ANOVA is a method for testing the hypothesis that three or more groups with normal distributions are the same. The sign test is similar to the t-test except that it uses the median rather than the mean. The F-test (aka: variance ratio test) is used to determine if standard deviations of two data sets are statically different. The F-test statistic is calculated by dividing the square of the larger variance by the square of the smaller variance. This value is evaluated using degree of freedom and an appropriate table.

4. a. The unpaired (student) t-test is used to compare two groups of observations that are unrelated to each other, such as diseased and non-diseased groups. The sample sizes do not have to be different. The paired t-test evaluates data derived from a single group (samples, patients) in which two different procedures or treatments are evaluated for differences.

5. e. With this change, fewer healthy individuals will be called positive. As you move to the right, the positive predictive value increases and negative predictive value decreases.

6. c. Choice C would likely demonstrate a shift or a drift. All others should demonstrate a Gaussian distribution. This distribution type is also known as a normal distribution wherein the population of data clusters symmetrically around a central value such that the mean, median, and mode of the data set are all identical.

7. d. CLIA-88 prohibits survey material to be shared among different facilities and laboratories. The PT survey should be used only for its intended facility and platform. Peer group evaluation allows a laboratory to verify that it is using a method according to manufacturer’s specifications. Furthermore, it demonstrates that patient results are consistent with those of other laboratories.

8. d. Sensitivity is defined as the percentage of individuals with a disease who have an abnormal result. Of the 100 children with the disease, 95 (95%) had an abnormal result.

9. e. Specificity is defined as percentage of normal results in individuals without the disease. Of the 300 children without the disease, 294 (300-6, or 98%) had a normal result.

10. b. When tested in a screening program, one must use the prevalence of disease to evaluate the predictive value. When calculating predictive value problems for the boards, approximations will allow you to determine the correct answer. If the test has 95% sensitivity, then it is likely that the one infant with disease will have a positive result. Of the 10,000 (actually 9,999) without disease, 2% would be expected to have an elevated value; this represents 200 individuals. The positive predictive value is calculated as the true positive results (1) divided by total positive results (201), multiplied by 100 to report it as a percent, or in this case 0.5%. It is not appropriate to use the 300 persons without disease and 100 persons with disease initially tested to calculate positive predictive value, because this does not represent the true frequency of disease in the population tested.

11. e. In an ROC curve, true positive rate or sensitivity is plotted on the y axis, and false-positive rate (100-specificity) is plotted on the x axis. In some versions of the plot, specificity is plotted on the x axis, but in that case the numbers on the x axis go from 100 on the left to 0 on the right.

12. c. In evaluating an ROC curve comparing more than one test, several features of the graph can be used to compare test performance against expectations. In viewing the graph, tests A and B reach sensitivity of approximately 100%; however, test A clearly has better specificity for any level of sensitivity chosen. In screening programs, sensitivity is the most important characteristic of test performance. On the other hand, test C has high specificity, but never reaches over approximately 80% sensitivity. Confirmatory tests require high specificity, so test C would be a good choice for a confirmatory test. Chance alone would be depicted by a straight line running from the lower left-hand corner to the upper right-hand corner of the graph. All of the tests perform significantly better than chance alone.

13. d. ROC curves are prepared by calculating sensitivity and specificity of a test at a number of different cutoff points. Inspection of the graphs is the best way to determine tests that may perform better as screening tests (high sensitivity) or confirmatory tests (high specificity). Area under the curve (i.e., the proportion of the square below and to the right of the ROC curve) is a common way to evaluate overall test performance; the higher the number (with 1.0 being a perfect test), the better. Since a straight line from the lower left to upper right corners (performance of chance alone) will include an AUC of 0.50, an AUC of 0.6 is only slightly better than chance alone. Efficiency refers to the proportion of all persons who are correctly classified by a test (true positive + true negative/total tested). Efficiency will be highest at that point closest to the upper left-hand corner of the graph.

14. d. (See discussion below in question 15.)

15. b. The t-test is a statistical evaluation of whether the means of two sets of observations (which have a gaussian distribution of results) are significantly different. There are several versions of the t-test, but the ones that cause the most confusion (and are often asked about on exams) are paired and unpaired t-tests. Basically, if the same samples or individuals are studied using two different drugs or two different tests, the paired t-test would be used. If different individuals or samples are studied, then the unpaired t-test would be used. The chi square test is typically used for evaluating differences in proportions between two groups. For example, consider a treatment for a disease. If 24/75 persons treated with drug A responded, while 10/57 treated with drug B responded, the response rates could be evaluated using the chi square test. The Kolmogorov-Smirnov test is used to determine whether a set of data follows a gaussian distribution.

16. b. According to the Clinical and Laboratory Standards Institute (CLSI), a laboratory is allowed to use a published reference interval, providing that it has validated that the reference interval is appropriate for its population. The approved protocol for performing this is to take 20 representative individuals and test their samples; if no more than two of the results are outside the published reference interval, it is said to be validated. If more than two are outside the interval, a second series of 20 can be tested, and the same criterion used to make sure the first observation was just by chance. Practically speaking, one would not wish to repeat this experiment if, say, half of the results were outside the published reference interval. If the published figures cannot be validated, then the laboratory must establish its own reference interval.

17. c. According to the CLSI standards, to achieve statistical reliability in determining limits for a reference interval, 120 individuals must be tested. The data should be inspected for outliers, and calculations repeated after excluding any such outlier data points.

18. d. In contrast to most other tests, upper reference limits for cholesterol and glucose are based on subsequent risk of developing clinical evidence of disease. By analyzing results of prospective studies, and analyzing test results at baseline, the values above which risk began to increase were chosen as the upper reference limits of healthy or desirable. In the case of cholesterol, this was at values markedly lower than would have been derived from studies of apparently healthy individuals, while for glucose it is at values markedly higher than from such studies.

19. c. For many tests, results within a person are kept within extremely narrow limits, while varying considerably from one person to the next. For such tests, population-based reference limits are insensitive to the presence of disease. Among common tests with small individual variation are alkaline phosphatase, calcium, and free thyroxine. On the other hand, individual variation is similar to population variation for the other tests listed here, along with phosphate.

20. c. This can be calculated similarly to calculating the odds that all 12 coin flips would turn up heads or tails. In that case, the likelihood of one flip resulting in a head is 0.5 (50%); the odds that all 12 would be heads is (0.5)12. Assuming all tests are independent variables, then the likelihood that all 12 would be within the reference interval is the likelihood of 1 (0.95) to the 12th power. The exact answer is 54%. This type of question commonly appears on the boards.

21. b. According to the CLSI guidelines on blood collection, the suggested order of collection is blood culture tubes first, followed (in order) by plain tubes, coagulation (sodium citrate) tubes, serum separator tubes, and tubes with additives. For plastic tubes, however, “plain” tubes often contain additives to accelerate coagulation; when plastic tubes are used, sodium citrate tubes should be filled before the “plain” tubes.

22. b. Prolonged use of a tourniquet causes hemoconcentration, which can increase the concentration of proteins and substances bound to proteins (calcium, iron, lipids, protein-bound hormones, and drugs).

23. a. While diurnal variation is best known for cortisol, there is significant diurnal variation for many other substances as well. Most pituitary hormone levels are lowest at bedtime and rise significantly during sleep to peak in the morning. Substances regulated by pituitary hormones (including many bone markers, such as osteocalcin, and collagen degradation fragments, such as N-telopeptides) also have marked diurnal variation. Glucose tolerance is affected by cortisol and growth hormone levels and is significantly better in the afternoon than in the morning.

24. b. During pregnancy, there is an increase in glomerular filtration rate of approximately 50%, resulting in a significant decrease in BUN and creatinine. It is important to consider this decrease in interpreting renal function tests during pregnancy. Increases in reference limits are needed for each of the other tests. Alkaline phosphatase levels increase to two-three times the nonpregnant reference limits, both from placental alkaline phosphatase and an increase in bone isoenzyme. During the second trimester, third spacing of fluid increases renin and aldosterone production. A number of hormones increase during pregnancy, including estriol (a marker of the fetal-placental unit) and prolactin.

25. a. Acute illness causes a number of changes in physiology. Cytokines lead to increased production of several pituitary hormones, including cortisol, growth hormone, and prolactin; however, gonadotropins are typically decreased. Increased cortisol causes increases in glucose and worsens glucose tolerance. Cytokines appear to increase uptake of cholesterol by macrophages; with acute illness, cholesterol often falls by up to 40%, and the decrease begins within 24 hours of onset of illness. Similar decreases occur in LDL-cholesterol, with lesser decreases in HDL-cholesterol, while triglycerides and VLDL increase by up to 35%.

26. d. Drug regimen and physical activity are more related to subject preparation. Site preparation and body posture are more related to specimen collection. Clotting is the only choice most likely related to specimen handling.

27. e. Heparin causes increases in laboratory testing such as thyroid hormone, PT, PTT, Li (Li-Heparin tubes), and sodium (Na-Heparin tubes). EDTA decreases iron; oxalate decreases calcium levels; citrate decreases amylase, and fluorides inhibit alkaline phosphatase activity.

28. a. C-peptide has about a 65% inter- and a 30% intraindividual CV. Calcium, glycohemoglobin, total protein, and glucose demonstrate at most ∼ 10% CV for either inter- or intraindividual CV.

29. d. Comparison studies are typically graphed using the current (or, ideally, a reference) method on the x axis and the new or comparison method on the y axis. Correlation coefficient is reflective of two things: the spread of points around the line and the range of values measured. High correlation coefficients occur when an adequate range of samples is tested and when the points are closely fit to the resulting line. Slope is a measure of proportional bias, while y intercept is a measure of constant bias.

30. a. Correlation coefficient is only a measure of fit of points to the line and an indicator of the range of the values measured. A perfect positive correlation would occur with an r value of 1, while a perfect negative correlation would be indicated by an r of − 1. A value of 0 indicates that there is no correlation between the two methods. The square of r (also called R), when expressed as a percentage, indicates the likelihood of predicting the value of y for a given x value. For method comparison studies, values of r should ideally be >0.97, and for most methods will be >0.99 unless there are significant problems with the comparison study. The square of the r value indicates how much of the individual value for y can be explained by the x variable; for method comparison studies, this should therefore be a very high number. The correlation coefficient does not indicate that values are equal; for example, a new method for an enzyme could give results three times those of the old method, but as long as that relationship always held, the r value would be high; as a consequence of this fact, correlation coefficient cannot be used to evaluate methods for bias.

31. b. Bias (Bland-Altman) plots are the most sensitive way to detect biases, particularly those that occur only in part of the measurement range. Bias plots can use either the absolute difference or the percent difference. When using either, the goal is to have the points centered on the zero line, with about half above and half below. Values that are consistently on one side of the zero line represent bias in that part of the measurement range.

32. b. Heterophile antibodies are a common cause of erroneous results in sandwich immunoassays, which are used to measure most tumor markers, peptide hormones, cardiac markers, and a number of other compounds. These antibodies react with immunoglobulins from other species, which may generate a signal in the immunoassay that is read as the presence of the analyte. While manufacturers take steps to minimize heterophile antibody interference, they vary in their effectiveness (even within different assays from the same manufacturer), so discrepant results are common in samples containing high titer heterophile antibodies. While fragments of hCG may also cause discrepant hCG results, they would not explain the positive PSA in a female, and ectopic PSA production has not been reported in trophoblastic tumors. Use of different calibrators may also cause slight differences between assays, but not to this degree.

33. d. Matrix effects are defined as differences in apparent concentration between samples with different composition (e.g., serum and urine). Matrix effects are common for substances that are poorly soluble in water (such as lipids) or in which there is a complex interaction between the sample and the reagents (such as enzyme assays and immunoassays). This creates problems in use of pure standards, which often have a very different matrix than serum samples. The solution to matrix effects is to use calibrators, solutions similar in composition to the samples being tested in which the concentration has been determined. Either calibrators or standards are used to prepare a standard (calibration) curve, in which the signal (e.g., amount of light absorbed) generated by the instrument on samples of known concentration is used to interpolate concentrations in unknown samples by measuring their generated signal in the assay.

34. b. The Beer-Lambert law is the basis for photometric measurement, which is the underlying principle for most chemical assays. It describes a linear relationship between amount of light absorbed and the concentration of a compound in a solution. While biologic samples have a variety of other compounds within them, use of chemical reactions specific for the compound of interest allows use of this relationship in the clinical laboratory. Transmittance has an inverse relationship to absorbance (light that was not absorbed is transmitted through a solution), so there is an inverse relationship between transmittance and concentration.

35. e. Within-run variation is a measure of repeatability of results over a single set-up of an instrument, typically during the course of a single shift, when a single operator sets up the instrument and performs a group of tests. The description given would apply to betweenrun variation, which is generally a larger number.

36. c. Coefficient of variation is a measure of the repeatability for a test, calculated as the percentage of the mean represented by the standard deviation of measurement. It is calculated by making multiple measurements of the same sample, either a control or a single patient. Coefficient of variation should be for electrolyte measurements, and in general is in the range of 2% to 5% for most tests. However, it may be as much as 10% for enzyme activity measurements, measurement of substances in very low concentrations, and for some immunoassays. For very complicated assays, it may be even higher. It is ideal that measurement variation be significantly lower than the variation within a person over time, so that changes in condition can be detected as changes in laboratory test results.

37. b. Turbidimetry and nephelometry are often used to measure the amounts of antigens (often proteins) that are present in concentrations of mg/dL or higher by measuring the number of antigen-antibody complexes present. The complexes form particles that prevent light from traveling in a straight line to a detector, instead causing the photons to be sent off at an angle to the incident light. This produces turbidity in a solution. Turbidimetry measures the decrease in light in the straight line path from the light source to the detector, while nephelometry places a detector at an angle to the incident light and measures the amount reaching the detector. Common applications are measurement of immunoglobulins, C-reactive protein, haptoglobin, and hemoglobin A1c. Anything else that causes turbidity in a sample, most commonly triglycerides, will interfere with measurement. Turbidimetry can be run using the same instrumental setup as spectrophotometry, which is used for most routine chemistry tests; nephelometry requires an instrument with a detector at an angle to the light source and is commonly done on dedicated instruments.

38. a. High-dose hook effect is seen in immunometric assays for substances that have very large ranges of concentration, most commonly tumor markers. When the range of assay concentration seen in samples exceeds 4-5 logs (for PSA, from <0.05 to over 5000, for example), there is the possibility of such an effect occurring. Since what is measured in immunometric assays is the number of antigen-antibody complexes, and this decreases with great antigen excess, a falsely low concentration results. Such samples can have extremely low apparent concentrations when the result is 6-7 logs above the usual measured concentration (e.g., for hCG measurements). PSA is not measured by its enzymatic activity, so those choices could not be correct. Heterophile antibodies typically cause falsely increased results, and would not be expected to disappear in the short period of time between diagnosis and first round of chemotherapy. While sample mix-up is always a possibility, it would be unusual to have another patient with the very high PSA value seen in the initial result.

39. a. Some reports indicate that as many as 50% of HAMAs are likely to dilute linearly. Therefore, the lack of proportional results with dilution provides useful information. However, proportional results alone are uninformative. Single dilutions should not be used to make decisions regarding the presence or absence of a heterophilic antibody. A more robust determination would consider a multipronged approach using blocking agents, measuring specifically for HAMA, and assaying the sample on more than one analytical platforms that use different antibodies.

40. d. NMR determines structures of organic compounds and, unlike MS, is not destructive. The other statements correctly describe each corresponding instrumentation principle.

41. b. The PCO2 electrode uses sodium bicarbonate buffer and a gas-permeable membrane. The calomel electrode is made of mercury that is in contact with a mercury (I) chloride-saturated solution. This solution is referred to as “calomel” and contains a known concentration of potassium chloride. The other choices correctly describe each of the corresponding electrodes.

42. c. Electroendosmosis describes the paradoxical migration of charged molecules (such as immunoglobulins) in a direction contrary to what would be expected. When support media with a negative surface charge (such as cellulose acetate or agarose) are used, hydronium ions (as H₃O+) are attracted to the surface and migrate toward the cathode (negatively charged) pole. The weak negative charge on proteins such as immunoglobulins causes them to be carried along with the solvent toward the cathode, instead of the expected migration toward the anode. Electroendosmosis does not occur with uncharged support media such as polyacrylamide gel or starch gel. It improves separation of proteins, and thus is helpful in analysis of serum proteins.

43. e. SDS-PAGE separates the denatured proteins primarily on the basis of molecular weight. Charge density is the most important factor in electrophoresis with nondenatured proteins. Degree of folding is irrelevant when proteins are denatured.

44. c. Temperature does not directly affect separation of compounds in electrophoresis; however, cooling is often used to prevent protein denaturation, which may affect migration of proteins. All of the other variables can be adjusted to improve separation of compounds in electrophoresis.

45. d. Whole blood would have to be centrifuged to assay serum or plasma. Centrifugation is not a feature consistent with the use of point-of-care testing devices. Sampling volumes are often smaller and are less traumatic for the patient, and built-in quality control checks lend to the ease of use.

46. d. Testing done on an analytical platform or bench-top analyzer would likely be a test performed in the main laboratory and would therefore not be a point-of-care test.

47. c. POCT analyzers generate more disposables, and reagent costs are usually more expensive than traditional laboratory systems. One of the main drawbacks of POCT systems is the expensive price.

48. c. In classic noncompetitive inhibition, the inhibitor binds only to the free enzyme and not to the enzyme-substrate complex. In uncompetitive inhibition, the inhibitor binds only to the enzyme-substrate complex and not to the free enzyme. Therefore, no enzymeinhibitor complex will form, and binding of the substrate will lead to a conformational change that will create an inhibitor binding site.

49. a. Acid phosphatase is unstable at all temperatures unless the pH of the serum is reduced to between 5 and 6.

50. d. Red cells do not contain creatine kinase. The highest ratio would be seen with lactate dehydrogenase (around 500:1). The ratio of AST is about 15:1, while that of ALT is about 7:1.

51. c. In contrast to other hepatocyte enzymes, pseudocholinesterase production by the liver reflects hepatic synthetic function rather than hepatocyte injury.

52. c. Very little interferes with screening tests for methadone. It is possible that methadone does not show up on confirmation because of the specific method used in the laboratory. Also, many reference laboratories such as the Mayo Clinic identify methadone separately, and it is not part of their screen to confirm opiates. Hydrocodone and hydromorphone are not metabolites nor do they interfere with methadone or oxycodone detection. Because both were found on gas chromatography mass spectrometry confirmation, a third drug is indicated. Methadone is likely the cause of the positive test result.

53. e. The legal limit for alcohol is 80 mg/dL (equivalent to “blowing” a 0.08 on a breathalyzer). Additionally, only whole blood (not serum) analyzed by head space gas chromatography from a chain-of-custody sample can be used to define a legal/forensic alcohol level. The driver’s measured level of 11 mg/dL (equivalent to blowing a 0.011) does not indicate impairment. Impairment is reported to begin at 40 mg/dL. Blood alcohol levels between 0.02 and 0.03 are associated with slight euphoria and loss of shyness but are not usually associated with loss of coordination. Depressant effects are not apparent at these levels. The presence of marijuana indicates past exposure and does not prove that an individual is under the influence of the drug. Urine immunoassays are screening tests, not confirmatory tests.

54. c. NaF is preferred for collecting samples from individuals who are suspected of cocaine use. Cocaine is metabolized by hydrolysis of ester linkages. In blood, cocaine is hydrolyzed to ecgonine methyl ester via cholinesterase. This reaction is dependent on the concentration of cocaine and may be inhibited by freezing or addition of fluoride or cholinesterase inhibitors. At 4°C 1 mg/L of cocaine will loose 100% of the parent drug after 21 days.

55. b. The dose is the amount of drug given. Bioavailability is the amount of an administered dose that enters the systemic circulation. The bioavailability of an intravenous dose is generally 100%. Volume of distribution is the apparent volume into which the administered drug has distributed. Clearance is the amount of serum completely cleared of drug in a unit of time (e.g., mL/min or L/hr). Steady state is achieved when the net rate of drug input is exactly matched by the rate of elimination. The rate going in equals the rate going out. The rate going out equals clearance. Steady state is the level achieved and maintained in the therapeutic range.

56. b. Diazepam (Valium) is a tranquilizer. Morphine is an opiate. Phenobarbital is a barbiturate. Cocaine is a stimulant (dopaminergic pathway). Phencyclidine is a hallucinogen.

57. c. Lidocaine is neither highly bound to protein nor is it appreciably stored in tissue. It has a relatively short half-life of about 2 hours. The other statements correctly state a function of the indicated cardiotropic agent.

58. e. About 90% of theophylline is metabolized by the liver, with caffeine being one of the major metabolites in children. In adults, caffeine is argued to be undetectable with theophylline use.

59. d. Opioids are associated with feelings of euphoria that are often accompanied by respiratory depression and bradycardia. Constipation and increased ADH secretion are other common associations with opioid ingestion.

60. a. In sweat, cocaine is primarily excreted as the parent drug. This offers an added advantage of simple gas chromatography analysis.

61. b. Aluminum phosphide is a grain fumigant that has been orally ingested in both suicides and accidents. Blood levels of aluminum rise during exposure and subsequently return to normal, unless kidney function is impaired. In increasing levels of toxicity, common forms of arsenic exist as arsenate (As⁺⁵), arsenite (As⁺³), and arsine (AsH₃) gas. Elemental arsenic is relatively not toxic. Effects of excess iron ingestion can include hepatic necrosis after 4 days and has also been associated with GI obstruction after 4 weeks. Patients taking Li₂CO₃ for manic-depressive disorders sometimes develop thyroid enlargement. Lead toxicity leads to inhibition of several mitochondrial enzymes in heme synthesis; ferrochelatase is directly affected.

62. e. Choices A-D all describe ADH-deficient causes of polyuria due to water diuresis. Choice E is not related to this pathogenesis.

63. e. Somatostatin inhibits a number of intestinal and pancreatic hormones. It causes diabetes through inhibition of insulin and hypochlorhydria due to inhibition of gastrin. Diarrhea is at least partially due to inhibition of pancreatic enzyme production, and gallstones may be due to inhibition of CCK production. While diarrhea can be seen with carcinoid and gastrinoma, the other findings are rare, while insulinomas only cause hypoglycemia and none of the other manifestations.

64. a. Measurement of plasma-free metanephrines is between 97% and 99% sensitive for diagnosing hereditary and sporadic pheochromocytomas and has the highest sensitivity among the choices listed for diagnosing a pheochromocytoma; however, specificity is relatively low (85%). Specificity is highest for measurement of fractionated metanephrines or catecholamines in urine.

65. c. Addison disease is associated with elevated levels of renin. Primary hyperaldosteronism and Cushing and Liddle syndromes are conditions associated with decreased renin levels. Additionally, decreased renin levels are associated with dexamethasone-suppressible hyperaldosteronism.

66. b. 11 beta-hydroxylase converts 11-deoxycortisol to cortisol in the glucocorticoid pathway (zona fasciculata) in the adrenal cortex. Congenital adrenal hyperplasia (CAH) associated with elevated levels of 11-deoxycortisol will likely be caused by a deficiency in 11 beta-hydroxylase. This enzyme defect represents approximately 5% of all cases of CAH, while 21-hydroxylase deficiency represents almost all of the remaining 95% of CAH cases. Much rarer causes include deficiencies of enzymes further “upstream” in the glucocorticoid and mineralocorticoid cascade.

67. e. A urine creatinine much less than 0.9 g/day likely means that the sample collection was incomplete. These values would lead to a misinterpreted catecholamine result.

68. c. The normal T4, T3 uptake, and TSH with low T3 are typical of “euthyroid sick syndrome,” or nonthyroidal illness. In acute illness, peripheral conversion of T4 to T3 (by the type I monodeiodinase) is impaired, while pituitary conversion of T4 to T3 (by the type 2 monodeiodinase) is normal, preventing an increase in TSH. Those who are acutely ill may actually have abnormal TSH due to the effects of a number of factors (cortisol, dopamine, cytokines) that impact TSH production, often making it hard to evaluate actual thyroid status in acute illness.

69. c.

Stay updated, free articles. Join our Telegram channel

Join