Clinical Chemistry I 0202304 PDF

Clinical Chemistry I 0202304 Instructor: Mohammad QABAJAH E-mail: [email protected] Week 03&04 Method Selection and Evaluation 2 Objectives I. To be familiar with the three major areas that aid them into the decision making with regard...

Clinical Chemistry I 0202304 Instructor: Mohammad QABAJAH E-mail: [email protected] Week 03&04 Method Selection and Evaluation 2 Objectives I. To be familiar with the three major areas that aid them into the decision making with regard to analytical method selection II. Understand the analytical performance of analytical methods III. Be aware of both the analytical performance and the practical performance criteria related to automated methods IV. Recall some of the statistical concepts related analytical performance of analytical methods. V. Able to compare between the analytical performance criteria using raw data by Bland-Altman method and correlation regression Method Selection ?? ?? 4 Method Selection: Considerations Medical Usefulness Analytical Performance Practical Criteria Patient Protocol Conditions Method Selection: Medical Usefulness - Achieving optimal patient care - What is needed clinically from a certain lab test Method Selection: Analytical Performance - Calibration - Precision - Accuracy (Trueness) - Analytical Range - Detection Limit - Clinical Sensitivity and Specificity Method Selection: Practical Criteria - The detailed protocol For automated procedures: - Reference materials - Pipetting precision - Reagents composition and stability - Carryover Specimen & Reagent - Technologist skills - Detector imprecision - Possible hazards and waste - Time to report - Specimen requirements - On-board reagent stability - Instrumental requirements - Overall throughput - Cost-effectiveness - Mean time of instrument failures - Computer platforms and interfacing - Mean time to repair - The availability of technical support Statistical Terms - Mean - Standard deviation - Coefficient of variation Gaussian Probability Distribution Analytical Methods: Basic Terms - Calibration - Linearity - Trueness /Accuracy - Precision - Limit of Detection Calibration The correlation between instrument signal and analyte concentration Y = Factor * X Instrument Signal - Y: Instrument Signal - X: Analyte Concentration Analyte Concentration Linearity: Relationship between measured and expected values over the range of analytical measurements Calibrator Known quantity Calibration: - Linear - Curved (immunoassays): Four-parameter-logistic curves Curved function: Nonlinear regression analysis is used or logit transformation Should be monotonic, otherwise, errors occur Stability of the signal &frequency of calibration Calibration - Random dispersion of instrument signal at a given concentration transforms into dispersion on the measurement scale - Modern automated machines: variations in response are very small, this means calibration is stable. Trueness and Accuracy Trueness: Closeness of agreement of the average measured value (high results #) with the True Value Accuracy: Closeness of agreement of a single measurement with True Value Trueness and Accuracy – Related Terms - Recovery: The difference between measured concentration and the amount added - Drif t: Differences Caused by instrument or reagent instability over time - Carryover: Differences caused by fraction of a samples measured with the next sample - Bias: mathematical difference between the average and true value  Precision - The closeness between independent results of measurements obtained under stipulated conditions - Imprecision: - Measured by SD or CV - Inversely related to precision - Caused by random error - Types - Between run precession - Interlaboratory pression  Precision – Related Terms - Repeatability: Closeness between results of successive measurements under the same conditions - Reproducibility: Closeness between results of successive measurements under changed conditions - How many measurement: The more observation the more the certainty - Commonly, 20 observations in duplicate Analytical Measurement Range and Limits of Quantification - Measuring Interval Reportable Range - The analyte concentration range over which measurements are within declared tolerances for imprecision and bias of the method Analytical Range Limit of Detection (LoD) & Limit of Blank (LoB) Important for many analytes especially hormones Factors Affecting: Instrument Sensitivity, Background Noise, Sample Matrix, Analyte Properties Errors: Type I error (NO analyte, signal present) Type II error (analyte present, NO signal) Reporting: Not detectable, less than LoD (zero, LoB) Detectable, greater than zero, detected Analytical Sensitivity - The ability of an analytical method to assess small variations in the concentration of analyte (Smallest concentration or amount of an analyte that can be measured and quantified with a high degree of accuracy and precision) - Depends on - Slope of the calibration curve (direct relationship) - Random variation (precision, CV) (inverse relationship) Analytical Specificity and Interference The ability of an assay procedure to determine specifically the concentration of the target analyte in the presence of potentially interfering substances or factors in the sample matrix Analytical Goals Based on: Clinical outcomes of clinical setting Biological variation Imprecision (σT)2 = (σ)2 within-B+ (σ)2A bias < 0.25 (σ2 within-B+ σ2between-B)0.5 Limits set by regulatory bodies Method Comparison Comparison study (points to be considered) - The number of samples necessary - The distribution of analyte concentrations - The representativeness of the samples - Practical aspects: Storage; treatment of samples; anticoagulants; measurement times - Ethical issues Comparison of Methods_Difference Bland-Altman Plot X1 X2 (X1 + X2)/2 X2-X1 (X2-X1)/((X2-X1)/2) 0.01005 99 100 99.5 1 190 200 195 10 0.051282 -0.03922 260 250 255 -10 Comparison of Methods_Regression Analysis - Estimate relationships between a dependent variable and one or more independent variables - Ordinary Least-Squares [OLS] - Beware of the outlier - https: //www.youtube.com/watch?v= P8hT5nDai6A X2 X1 100 99 200 190 250 260 Comparison of Methods_Proportional random error - Depends on the change in a specific variable (directly related) - Measurable amount: X divided by Y always equals the same constant. Comparison of Methods_Traceability Unbroken chain of comparisons of measurements leading to a known reference value the property of the result of a measurement or the value of a standard whereby it can be related to stated reference through unbroken chain of comparisons (chain of calibration leading to primary national or international standards) Comparison of Methods_Uncertainty A parameter associated with the result of a measurement that characterizes the dispersion of the values. Expressed by SD The location of true value with a given probability ~ 95 % Affected by - Preanalytical variation - Method imprecision - Sample-related random interferences - Uncertainty related to calibration - Bias corrections (traceability) See YOU Next Lecture ☺ Clinical Chemistry I 0202304 Instructor: Mohammad QABAJAH E-mail: [email protected] Week 05&06 Clinical Evaluation of Laboratory Tests 2 Objectives I. State the formulas used in calculating clinical sensitivity, clinical specificity, and predictive value of a laboratory test. II. State how the predictive value of a laboratory test is affected by prevalence. III. Construct a receiver operating characteristic plot using data from a diagnostic test study. IV. Interpret the difference plot used in method comparison. Agreement of test results: Accurate diagnosis - Sensitivity and Specificity - Predictive Values - Prevalence - Likelihood Ratio (LR) - Odds Ratio (OR) Review: Gaussian Probability Distribution Sensitivity and Specificity Negative and Positive Source of the False  Reference Method (Gold/Reference Standard) “Best current practice for establishing the presence of a disorder” I. Screening; High sensitivity: (high FP, low FN) ̵ Pick up all positive results for further diagnosis ̵ Make sure nobody is missed II. Diagnostic; High specificity: (low FP, high FN) ̵ Exclude all the people who do not have the disease ̵ Make sure nobody is picked Screening Vs Diagnostic: Run the diagnostic test on all positive screened Sensitivity and Specificity # Patients with # Patients with Positive test result Negative test result # Patients with disease True Positive (TP) False Negative (FN) # Patients without disease False Positive (FP) True Negative (TN) Sensitivity and Specificity Sensitivity: The proportion of subjects with disease who have a positive laboratory test result 𝑇𝑃 𝑇𝑃 Sensitivity = 𝑇𝑃+𝐹𝑁 , % Sensitivity = 𝑇𝑃+𝐹𝑁 X 100 # Patients with # Patients with Positive test result Negative test result # Patients with disease True Positive (TP) False Negative (FN) # Patients without disease False Positive (FP) True Negative (TN) Sensitivity and Specificity Specifi city: The proportion of subjects without disease who have a negative laboratory test result. 𝑇𝑁 𝑇𝑁 Specifi city = 𝐹𝑃+𝑇𝑁 , % Specifi city = 𝐹𝑃+𝑇𝑁 X100 # Patients with # Patients with Positive test result Negative test result # Patients with disease True Positive (TP) False Negative (FN) # Patients without disease False Positive (FP) True Negative (TN) IF IF Spec. = 95% Sen. = 90% # of tested -ve = 100 # of tested +ve = 100 TN ? TP ? FP ? FN ? Predictive Values Positive Predictive Value (PV+): Negative Predictive Value (PV-): The fraction with a positive test who The fraction with a negative test who have the disease don’t have the disease 𝑇𝑃 𝑇𝑁 𝑃𝑉+ = 𝑃𝑉− = 𝑇𝑃 + 𝐹𝑃 𝑇𝑁 + 𝐹𝑁 # Patients with # Patients with Positive test result Negative test result # Patients with disease True Positive (TP) False Negative (FN) # Patients without disease False Positive (FP) True Negative (TN)  Prevalence: The frequency of disease in the population examined Screening and prevalence Populat ion 1,000,000 Prevalence 1 in 10 (0.1) People wit h disease 100,000 (0.1 X 1,000,000) Example: Sensitivity, Specificity, PV+, and PV- Calculate: Sensitivity, Specificity, PV+, and PV- Positive Result Negative Result Have disease 100 5 Don’ t have disease 7 95 Likelihood Ratio (LR) The Likelihood Ratio (LR) is the likelihood that a given test result would be expected in a patient with the target disorder compared to the likelihood that that same result would be expected in a patient without the target disorder. Positive Likelihood Ratio (+LH) = Sensitivity/(1-Specificity) Negative Likelihood Ratio (-LH) = (1- Sensitivity)/Specificity Likelihood Ratio (LR) Test is GOOD if : +LH > 10; -LH < 0.1 19 20 21 Odds Ratio (OR) The probability of the presence of a specific disease divided by the probability of its absence (+LH) = Sensitivit y/(1-Specifi city) (-LH) = (1- Sensitivit y)/Specifi city 24 25 See YOU Next Lecture Clinical Chemistry I 0202304 Instructor: Mohammad QABAJAH E-mail: [email protected] Week_07 Reference Values 2 Objectives I. State the need for establishing reference intervals for the major clinical tests. II. List conditions that are essential when a valid comparison of individual laboratory results with reference values is performed III. Understand the exclusion criteria used in the production of health- associated reference values IV. List the partitioning criteria used to subgroup a reference group V. state the parametric and nonparametric statistical methods of determining an interpercentile interval VI. Discuss the issue of transferability of reference values with regard to prerequisites and solutions to the issue. VII. Compare the terms reference value and reference interval Decision Making: Comparison Medical interviews, Clinical examinations, Supplementary investigations ??? Does not resembles resembles that of a that of a typical typical particular particular disease disease Exclude diagnosis Diagnose In laboratory: Decision Making by Comparison Reference values ̵ Healthy Individuals ̵ Patients with Diseases Reference values; Obtained from ̵ single individual (Subject-based reference values) ̵ or group of individuals (Population-based reference values) Corresponding to a stated description, which must be spelled out and made available for use by others Comparing of Observed Data Several collections of reference values - Healthy people - Undifferentiated hospital population Reference Individuals - People with typical diseases Conditions to Ensure Proper Comparison - All groups of reference individuals should be clearly defined - The patient examined should resemble sufficiently - The conditions should be known - All quantities should be of the same type - Standardized methods for all results - The clinical sensitivity, clinical specificity, and prevalence should be known Selection of Reference Individuals Ideally: the whole population Random sampling (as large as possible (120~ Exclusion Criteria - Diseases - Risk Factors - Obesity - Hypertension - Risks from occupation or environment - Genetically determined risks - Intake of Pharmacologically Active Agents - Specific Physiological States - Pregnancy - Stress - Excessive exercise Subgrouping of the Reference Group - Age - Gender - Physiological Factors - Stage in menstrual cycle - Stage in pregnancy - Physical condition Specimen Collection Preanalytical standardization ̵ Preparation of individuals before sample collection ̵ The sample collection ̵ Handling of the sample before analysis Analytical Procedures and Quality Control Components that need to be defined - Analysis method, including information on - Equipment - Reagents - Calibrators - Types of raw data - Calculation method - Quality control - Reliability criteria Statistical Treatment of Reference Values - Partitioning of the reference values into appropriate groups - Inspection of the distribution of each group - Determination of reference limits - Identification of outliers Inspection of Distribution Display the reference distribution graphically (frequency distribution) - Unimodal, Bimodal or polymodal distribution - The shape of the distribution - Highly deviating values (outliers) Determination of Reference Limits Reference interval: ̵ Upper reference limit and a lower reference limit ̵ Representing specified proportion of the reference population 95% Range: (Higher value - Lower value) Clinical decision limits: ̵ Provide optimal separation among clinical categories. (healthy individuals, patients with relevant diseases) Identification of outliers Assumption that the distribution is Gaussian An outlier is defined as being any point of data that lies over 1.5 IQRs below the first quartile (Q1 :25th percentile) or above the third quartile (Q3: 75th percentile) in a data set. 19 Determination of Reference Limits ◦ Reference limit ◦ Clinical decision limit ◦ Reference range vs. Reference interval ◦ Reference interval: Tolerance interval Interpercentile interval Prediction interval Interpercentile Interval ◦Simple to estimate, most used; recommended by IFCC, used for: - Parametric techniques: Calculation based on mean and SD 2.5 Percentile = Mean - 1.96 SD 97.5 Percentile = Mean + 1.96 SD Lower limit = Percentile Limit – 2.81*SD/√n Upper limit = Percentile Limit + 2.81*SD/√n - Non-parametic techniques: Makes no assumption about the data ◦Based on the rank ◦Sort data, compute rank numbers 2.5 and 97.5 ◦Find reference value corresponding to the rank ◦Determine the confidence interval 2.5% 97.5% Clinical Decision Limit_ Example Transferability/Transference of Reference Values The use of reference values generated by other laboratory Major requirements (population wise) ̵ Populations must be comparable ̵ No major ethnic differences ̵ No major social differences ̵ No major environmental differences Transferability/Transference of Reference Values Major requirements (pre analytical wise) ̵ Qualifying reference individuals ̵ Preparing individuals for specimen collection ̵ Performing specimen collection ̵ Using the same analytical method Transferability/Transference of Reference Values Points to be considered (analytical wise) ̵ Standardization of analytical protocols ̵ Common calibration ̵ Design of a sufficiently efficient external quality control scheme ̵ Mathematical transferring if results are not directly comparable Transferability/Transference of Reference Values Verification (CLSI recommendation) ̵ Measure 20 individuals ̵ If no more than 2 outside values, adopt reference values ̵ Multicenter production of reference values: Minimum 120 readings See YOU Next Lecture ☺