Harr MLS Review Chemistry 5.4 Calculations, Quality Control, and Statistics
A new method for BUN is evaluated by comparing the results of 40 paired patient samples to the urease-UV method. Normal and high controls were run on each shift for 5 days, five times per day. The results are as follows: Linear regression = ŷ = -0.3 + 0.90x Low control: x = 14.2 mg/dL; s = 1.24 High control: x = 48.6 mg/dL; s = 1.12 What is the total analytical error estimate for a sample having a concentration of 50 mg/dL?
-7.5 mg/dL Linear regression analysis gives an estimate of SE, which is equal to (ŷ - xc) where xc is the expected concentration, and ŷ is the value predicted by the linear regression equation. SE = [-0.3 + (0.9 × 50 mg/dL)] - 50.0 mg/dL = 44.7-50.0 = -5.3 mg/dL The standard deviation of the new method for the high control is used to estimate the RE because the mean of this control is nearest to the expected concentration of 50 mg/dL. RE is estimated by ±1.96 × s. RE = 1.96 × 1.12 = ± 2.2 mg/dL Total analytical error (TE) is equal to the sum of SE and RE. TE = SE + RE = -5.3 mg/dL + (-2.2 mg/dL) = -7.5 mg/dL
Which of the following quality control (QC) rules would be broken 1 out of 20 times by chance alone?
1 2S
What is the pH of a solution of HNO3, if the hydrogen ion concentration is 2.5 × 10^-2 M?
1.6
Two methods for total cholesterol are compared by running 40 paired patient samples in duplicate on each instrument. The following results are obtained: Instrument: Method x (reference method) Mean: 235 mg/dL Standard Deviation: 3.8 Instrument: Method y (candidate method) Mean: 246 mg/dL Standard Deviation: 3.4 Assuming the samples are collected and stored in the same way and the analysis done by a technologist who is familiar with both methods, what is the bias of method y?
11.0
A procedure for aspartate aminotransferase (AST) is performed manually because of a repeating error code for nonlinearity obtained on the laboratory's automated chemistry analyzer; 0.05 mL of serum and 1.0 mL of substrate are used. The reaction rate is measured at 30°C at 340 nm using a 1.0 cM light path, and the delta absorbance (-∆A) per minute is determined to be 0.382. Based upon a molar absorptivity coefficient for NADH at 340 nm of 6.22 X 103 M^-1 cM^-1 L^-1, calculate the enzyme activity in international units (IUs) per liter.
1290 IU/L
How many milliliters of HNO3 (purity 68.0%, specific gravity 1.42) are needed to prepare 1.0 L of a 2.0 N solution? Atomic weights: H = 1.0; N = 14.0; O = 16.0
130.5 mL
A glycerol kinase method for triglyceride calls for a serum blank in which normal saline is substituted for lipase in order to measure endogenous glycerol. Given the following results, and assuming the same volume of sample and reagent are used for each test, calculate the triglyceride concentration in the patient's sample. Standard concentration = 125 mg/dL Absorbance of Reagent Blank = 0.000 Absorbance of Standard = 0.62 Absorbance of Patient Serum = 0.750 Absorbance of Serum Blank = 0.100
131 mg/dL
A biuret reagent requires preparation of a stock solution containing 9.6 g of copper II sulfate (CuSO4) per liter. How many grams of CuSO4 • 5H2O are needed to prepare 1.0 L of the stock solution? Atomic weights: H = 1.0; Cu = 63.6; O = 16.0; S = 32.1
15.0 g
A new test for prostate cancer is found to have a sensitivity of 80.0% and a specificity of 84.0%. If the prevalence of prostate cancer is 4.0% in men over 42 years old, what is the predictive value of a positive test result (PV+) in this group?
17.2%
A procedure for cholesterol is calibrated with a serum-based cholesterol standard that was determined by the Abell-Kendall method to be 200.0 mg/dL. Assuming the same volume of sample and reagent are used, calculate the cholesterol concentration in the patient's sample from the following results. Standard concentration = 200 mg/dL Absorbance of Reagent Blank = 0.00 Absorbance of Standard = 0.860 Absorbance of Patient Serum = 0.740
172 mg/dL
Two consecutive controls are both beyond -2s from the mean. How frequently would this occur on the basis of chance alone?
1:1,600
How many significant figures should be reported when the pH of a 0.060 M solution of nitric acid is calculated?
2
What is the pH of a buffer containing 40.0 mmol/L NaHC2O4 and 4.0 mmol/L H2C2O4? (pKa = 1.25)
2.25
Convert 2.0 mEq/L magnesium (atomic weight = 24.3) to milligrams per deciliter.
2.4 mg/dL
Convert 10.0 mg/dL calcium (atomic weight = 40.1) to International System of Units (SI).
2.5
A solvent needed for HPLC requires a 20.0 mmol/L phosphoric acid buffer, pH 3.50, made by mixing KH2PO4 and H3PO4. How many grams of KH2PO4 are required to make 1.0 L of this buffer? Formula weights: KH2PO4 = 136.1; H3PO4 = 98.0; pKa H3PO4 = 2.12
2.61 g
How many milliliters of glacial acetic acid are needed to prepare 2.0 L of 10.0% v/v acetic acid?
200.0 mL
What is the pH of a 0.05 M solution of acetic acid? Ka = 1.75 × 10^-5, pKa = 4.76
3.0
Given the following data, calculate the coefficient of variation for glucose. Analyte: Glucose Mean: 76 mg/dL Standard deviation: 2.3
3.0%
The following plot represents a study of a screening test for malignant prostate cancer using plasma PSA (ng/mL). The outcome measured was positive cytology results obtained by biopsy. What concentration gives the highest sensitivity with the least number of unnecessary biopsies?
3.6 A receiver operating characteristic (ROC) curve is used to identify the test result, giving the highest sensitivity with the least number of false-positive results. Sensitivity (true positives) is plotted against false positives. The number in the uppermost left corner represents the highest detection with the lowest number of false positives. In this case, a result of 3.6 ng/mL detects 72% of malignancies with 1 in 10 (10%) false positives.
How many grams of sodium hydroxide (NaOH) are required to prepare 150.0 mL of a 5.0% w/v solution?
7.5 g
How many milliliters of a 2,000.0 mg/dL glucose stock solution are needed to prepare 100.0 mL of a 150.0 mg/dL glucose working standard?
7.5 mL
A new tumor marker for ovarian cancer is evaluated for sensitivity by testing serum samples from patients who have been diagnosed by staging biopsy as having malignant or benign lesions. The following results were obtained: Number of malignant patients who are positive for CA 125 = 21 out of 24 Number of benign patients who are negative for CA 125 = 61 out of 62 What is the sensitivity of the new CA 125 test?
87.5% Sensitivity is defined as the percentage of persons with the disease who have a positive test result. It is calculated as true positives (TP) divided by the sum of TP and false negatives (FN). % Sensitivity = [TP × 100]/[TP + FN] Sensitivity = (21 × 100) ÷ (21 + 3) = 87.5%
Calculate the pH of a solution of 1.5 × 10^-5 M NH4OH.
9.2
All of the following are requirements for a QC material except:
Analyte concentration must be independent of the method of assay
Which of the following assays has the poorest precision?
Analyte: Ca Mean (mmol/L): 2.5 Standard deviation: 0.3
In which circumstances is a validation study (versus performing routine quality control) required?
Change in reagent lot All of the listed conditions except a change in the reagent lot number can be validated by assaying two levels of control material following the procedure. A change in reagent lot number may alter the test system more dramatically, especially when the reagent was subjected to storage and shipping conditions that alter its performance. Therefore, both controls and patient samples should be analyzed and the results compared to the reagent in current use using criteria determined by the total allowable error for the analyte.
Which plot will give the earliest indication of a shift or trend?
Cusum
Given the following QC chart, identify the day in which a violation of the R4s QC rule occurs.
Day 15 An R4s error is defined as the algebraic difference between two controls within the same run. In this Levy-Jennings plot, on day 15, Level 1 is above the +2s limit (approximately +2.5s) and Level 2 is below the -2s limit (approximately -2.5s). These controls are approximately 5s apart (+2.5s minus -2.5s = +5s).
Referring to the Levy-Jennings chart, what is the first day in the second half of the month that patient results would be rejected?
Day 17 The 4 1s rule is broken across QC levels on day 17. This means that four consecutive controls are greater than ±1s from the mean. QC rules that are sensitive to SE are applied across both runs and levels to increase the probability of error detection. These are 22s, 41s, and 10Xbar.
Referring to the Levy-Jennings chart, what is the first day in the month when the run should be rejected and patient results should be repeated?
Day 8 Although the trend is apparent across QC levels by day 7, the patient results would not be rejected until day 8 when the 41s rule is broken. An advantage to plotting control data is that trends can be identified before results are out of control and patient data must be rejected. In this case, corrective steps should have been implemented by day 7 to avoid the delay and expense associated with having to repeat the analysis of patient samples.
Which of the following methods is most useful in order to detect sample misidentification?
Delta limit Cumulative summation is a statistical method used in quality assurance to detect a trend in QC results. Critical limits are used to define when medical intervention is likely needed to prevent injury or death. The significant change limit is the difference in test results that is medically significant, or that which cannot be attributed to the sum of normal physiological and analytical variation. The delta limit (delta check) determines whether the difference between two measurements usually 24-48 hours apart exceeds the expected. Although this can result from an abrupt change in the patient's status, other causes are sample misidentification, contamination, and random error. Delta limits are expressed in percent and vary depending on analyte stability.
A trend in QC results is most likely caused by:
Deterioration of the reagent
When establishing QC limits, which of the following practices is inappropriate?
Exclusion of any QC results greater than ±2s from the mean
What measurement in addition to true negatives and prevalence is required to calculate the predictive value of a negative test result (PV-)?
False negatives The PV- is defined as the probability that a person with a negative test result is free of disease. A high PV- is a characteristic of a good screening test. The predictive value of a negative test is calculated by multiplying the true negatives by 100, then dividing by the sum of the true negatives and false negatives. %PV- = [TN × 100]/[TN + FN]
In addition to the number of true negatives (TN), which of the following measurements is needed to calculate specificity?
False positives The clinical specificity of a laboratory test is defined as the true negatives divided by the sum of true negatives and false positives (FP). % Specificity = [TN × 100]/[TN + FP] Specificity is defined as the percentage of disease-free people who have a negative test result. The probability of false positives is calculated from the specificity as: 1-(% specificity/100)
Which of the following plots is best for detecting all types of QC errors?
Levy-Jennings
A laboratory is establishing a reference range for a new analyte and wants the range to be determined by the regional population of adults age 18 and older. The analyte concentration is known to be independent of race and gender. Which is the most appropriate process to follow?
Measure the analyte in 120 healthy adults and calculate the central 95th percentile Since the concentration of an analyte may not be normally distributed in a population, the reference range should not be determined from the standard deviation. It is more appropriate to determine the central 95th percentile (the range that encompasses 95% of the results). A minimum of 120 samples is needed for statistical significance. Results are rank ordered from lowest to highest. The 3rd result is the lowest value and the 118th is the highest value in the reference range. The laboratory can verify a preexisting reference range (e.g., as determined by the manufacturer's study) by testing 20 healthy persons. If no more than 10% fall outside the range, it can be considered valid for the patient population.
One of two controls within a run is above +2s and the other control is below -2s from the mean. What do these results indicate?
Poor precision has led to random error (RE)
When the magnitude of error increases with increasing sample concentration, it is called:
Proportional error Proportional error (slope or percent error) results in greater absolute error (deviation from the target value) at higher sample concentration. Constant error refers to a difference between the target value and the result, which is independent of sample concentration. For example, if both level 1 and level 2 controls for laboratory A average 5 mg/dL below the cumulative mean reported by all other laboratories using the same method, then laboratory A has a constant error of -5 mg/dL for that method.
In most circumstances, when two controls within a run are both greater than ±2s from the mean, what action should be taken first?
Recalibrate, then repeat controls followed by selected patient samples if quality control is acceptable
Referring to the Levy-Jennings chart, what analytical error is present during the second half of the month?
Shift A shift is characterized by six consecutive points lying on the same side of the mean. This occurs from day 15 to day 20. Shifts are caused by a change in the assay conditions that affect the accuracy of all results, such as a change in the concentration of the calibrator; change in reagent; a new lot of reagent that differs in composition; or improper temperature setting, wavelength, or sample volume. The term kurtosis refers to the degree of flatness or sharpness in the peak of a set of values having a Gaussian distribution.
Which of the following total quality management tools can be used to calculate the analytical error rate for an analyte in the clinical laboratory?
Six sigma All four of the answer choices are total quality management (TQM) tools used in the clinical laboratory to improve performance. Six sigma is a measurement of the frequency of product defects. In clinical laboratories, it refers to the frequency of an erroneous result. At the six-sigma level, the analytical process has such small variance that an error of six times the standard deviation would still be within acceptable limits for total allowable error. For example, a six-sigma process for an analyte produces a significant error in test result only 3.4 out of 1 million times the test is performed. Conversely, a method performing at the three-sigma level would give 66,807 errors per million. The sigma of a method is calculated by subtracting its bias from the total method error and dividing by its standard deviation. It is the method's random error divided by its standard deviation.
When comparing the laboratory's monthly mean to its peer group to determine if bias is present, what statistic is most appropriate?
Standard deviation index The standard deviation index (SDI) compares the lab's mean to the peer group's mean in terms of standard deviations instead of concentration. This normalizes the value so that it is independent of mean, and allows performance comparisons for any analyte. The SDI equals the lab's mean minus the peer group's mean divided by the peer group's standard deviation. It has a similar probability distribution to a t test and a value greater than 2.0 is considered significant.
Which of the following statistical tests is used to compare the means of two methods?
Student's t test Student's t test is the ratio of mean difference to the standard error of the mean difference (bias/random error) and tests for a significant difference in means. The F test is the ratio of variances and determines if one method is significantly less precise. The correlation coefficient is a measure of the association between two variables and should be high in any method comparison. An r value less than 0.90 in method comparisons usually occurs when the range of results is too narrow.
When referring to quality control (QC) results, what parameter usually determines the acceptable range?
The 95% confidence interval for the mean
Which explanation is the best interpretation of the following BUN bias plot?
The new method is lower than the reference and the magnitude is concentration dependent A bias plot compares the bias (candidate method minus reference method) to the result of the reference method. Ideally, points should be scattered equally on both sides of the zero line. When the majority of points is below the zero line, the candidate method is negatively biased (lower than the reference). In this case, the difference between the methods increases in proportion to the BUN concentration. This type of plot occurs when the slope of the linear regression line is low.
Serum samples collected from hospitalized patients over a 2-week period are split into two aliquots and analyzed for prostate specific antigen (PSA) by two methods. Each sample was assayed by both methods within 30 minutes of collection by a technologist familiar with both methods. The reference method is method × (upper reference limit = 4.0 μg/L). Linear regression analysis was performed by the least-squares method, and results are as follows: Linear regression: ŷ = 2.10 + 1.01x Correlation coefficient (r) = 0.984 Standard Error of Estimate (s y/x) = 0.23 Which statement best characterizes the relationship between the methods?
There is a significant bias caused by constant error The linear regression analysis is the most useful statistic to compare paired patient results because it estimates the magnitude of specific errors. The y intercept of the regression line is a measure of constant error, and the slope is a measure of proportional error. Together, these represent the bias or SE of the new method. The correlation coefficient is influenced by the range of the sample and the RE. Two methods that measure the same analyte will have a high correlation coefficient, provided the concentrations are measured over a wide range, and this statistic should not be used to judge the acceptability of the new method. The standard error of estimate is a measure of the closeness of data points to the regression line and is an expression of RE.
Which statement best summarizes the relationship between the new BUN method and reference method based upon the following linear regression scatterplot?
There is little or no constant error, but some proportional error The scatterplot shows that each sample produces a coordinate (x corresponds to the reference result and y to the candidate method result) that is very close to the regression line. This means that the variance of regression is low and there is a high degree of certainty that the predicted value of y will be close to its measured value. Near-zero concentration there is good agreement between methods; however, the higher the result, the greater the difference between x and y. The regression equation for this scatterplot is y = -0.01 + 0.90 x, indicating a proportional error of -10%.
Two freezing point osmometers are compared by running 40 paired patient samples one time on each instrument, and the following results are obtained: Instrument: Osmometer A Mean: 280 mOsm/kg Standard Deviation: 3.1 Instrument: Osmometer B Mean: 294 mOsm/kg Standard Deviation: 2.8 If the critical value for F = 2.8, then what conclusion can be drawn regarding the precision of the two instruments?
There is no statistically significant difference in precision
Which of the following plots is best for comparison of precision and accuracy among laboratories?
Tonks-Youden
Examine the Levy-Jennings chart and identify the QC problem that occurred during the first half of the month.
Trend A trend is characterized by six consecutive decreasing or increasing control results. The value for both controls becomes progressively higher from day 4 to day 9. Trends are caused by changes to the test system that increase over time, such as deterioration of reagents or calibrators, progressive changes in temperature, evaporation, light exposure, and bacterial contamination. A trend is a type of SE because all results are affected. Conversely, RE affects some results (but not others) in an unpredictable manner. Control rules affected by RE are 13s and R4s.
Which of the following conditions is cause for rejecting an analytical run?
Two consecutive controls greater than 2 s above or below the mean
The term R4S means that:
Two controls in the same run are greater than 4s units apart
What is the minimum requirement for performing QC for a total protein assay?
Two levels assayed within 24 hours The minimum requirement for frequency of quality control for a general chemistry analyte (based upon the Clinical Laboratory Improvement Act, 1988) is two levels of control assayed every 24 hours. Some laboratories prefer to assay two control levels every 8 hours to increase the opportunity for error detection. Two controls every 8 hours are required for blood gases, automated hematology, and point-of-care glucose testing to comply with College of American Pathology requirements. Analytes that display different CVs at the low, normal, and high ranges require 3 levels of control in 24 hours. These include blood gases, therapeutic drugs, and hormones.
