Laboratory Mathematics and Diagnostic Performance

Significant figures express the precision used to report a measured value. The number of digits does not quantify measurement uncertainty, which requires a separate estimate. Only measured quantities limit the reported precision of a calculation. Exact counts and defined conversion factors are exact.

For routine textbook calculations, addition and subtraction follow the least precise decimal place among the measured values. Multiplication and division follow the measured value with the fewest significant figures. A measurement procedure or reporting policy may specify a different rule.^{1,21 Thompson A, et al. Guide for the Use of the International System of Units (SI). 2008.2 Bishop ML, et al. Clinical Chemistry. 2023.}

• 9.7 mL + 2.84 mL + 0.316 mL = 12.856 mL, reported as 12.9 mL because 9.7 mL reaches the tenths place.
• 6.2 mg/mL × 3.45 mL = 21.39 mg, reported as 21 mg because 6.2 has two significant figures.

Carry unrounded values through intermediate steps. These examples use round-to-nearest with ties to even when the discarded portion is exactly 5 followed only by zeros: 4.85 → 4.8 and 3.75 → 3.8 at two significant figures. A nonzero digit after the 5 makes the discarded portion greater than half and rounds the retained digit upward. Laboratory procedures and information systems may specify another tie-breaking rule. Scientific notation makes the intended precision of trailing zeros clear, as in 3.65 × 10³ mg/dL.

A proportion states that two ratios are equal. Match the units before cross-multiplying and solving for the unknown. A reagent formula contains 2.00 g per 100.0 mL, so 2.00 g/100.0 mL = x g/300.0 mL. Cross-multiplication gives 100.0x = 600.0, or x = 6.00 g. Writing the unit beside each number makes an inverted conversion or mismatched volume easier to spot before the arithmetic begins.

A dilution contains a measured volume of sample and a diluent. Ratio notation differs among methods, so always state what the denominator is.^{2,32 Bishop ML, et al. Clinical Chemistry. 2023.3 Clinical and Laboratory Standards Institute. Establishing and Verifying an Extended Measuring Interval… 2018.}

Dilution fraction = sample volume ÷ final volume

Dilution factor = final volume ÷ sample volume

In a 1-in-8 dilution with a final volume of 320 µL, each part is 320 ÷ 8 = 40 µL. Prepare it with 40 µL sample and 280 µL diluent.

Correct the diluted specimen result with the factor that matches the prepared sample and diluent volumes:

Original concentration = diluted result × dilution factor

Mixing 25 µL serum with 225 µL diluent gives 250 µL total and a dilution factor of 10. A diluted triglyceride result of 365 mg/dL corresponds to 365 × 10 = 3.65 × 10³ mg/dL in the original specimen. For patient testing, use a diluent, ratio, and measuring interval supported by the manufacturer or validated by the laboratory. The Dilution Math and Dilution station pages provide more practice with preparation and correction.

For a serial dilution, multiply the factor from every completed step. Three successive 1-in-5 dilutions produce a cumulative dilution of 1/5 × 1/5 × 1/5 = 1/125 and a cumulative factor of 125. Using 10 µL transferred sample and 40 µL fresh diluent at each step consumes 10 µL of original specimen and 120 µL of fresh diluent across the series. A 1-in-15 step followed by four twofold steps gives a cumulative factor of 15 × 2⁴ = 240, or a 1-in-240 total dilution.

A titer is the reciprocal of the highest tested dilution that meets the assay's defined positive endpoint. In a twofold series from 1/2 through 1/512, an endpoint present through 1/64 and absent from 1/128 onward gives a reported titer of 64. The assay procedure defines the endpoint, reaction conditions, and reporting form.

Molarity and normality appear in the examination outline and in many laboratory instructions. The SI unit for amount-of-substance concentration is mol/m³, although clinical laboratories commonly use mol/L. Normality is a reaction-specific legacy convention expressed in eq/L.^{11 Thompson A, et al. Guide for the Use of the International System of Units (SI). 2008.}

Molarity

Calculate amount in moles from mass and molar mass:

Amount (mol) = mass (g) ÷ molar mass (g/mol)

Amount concentration (mol/L) = amount (mol) ÷ final solution volume (L)

The traditional symbol M denotes mol/L, as in 1.25 M. To prepare 600 mL of 1.25 mol/L NaOH with a molar mass of 40.00 g/mol:

Mass = 1.25 mol/L × 0.600 L × 40.00 g/mol = 30.0 g NaOH

Dissolve the measured solute and bring the solution to its stated final volume.

Normality

Normality is expressed in equivalents per liter and depends on the reaction being described. The n-factor counts the protons, hydroxide ions, electrons, or charge units that participate in that reaction.^{22 Bishop ML, et al. Clinical Chemistry. 2023.}

Equivalent mass (g/eq) = molar mass (g/mol) ÷ reaction-specific n-factor (eq/mol)

Normality (eq/L) = mass (g) ÷ [equivalent mass (g/eq) × solution volume (L)]

Normality = n-factor × molarity

For complete neutralization of both protons in H₂SO₄, the n-factor is 2. Its equivalent mass is 98.08 ÷ 2 = 49.04 g/eq. Calculate the amount needed to prepare 850 mL of 2.40 N H₂SO₄:

2.40 eq/L × 0.850 L × 49.04 g/eq = 1.00 × 10² g H₂SO₄

The n-factor depends on the stated reaction:

Percent solutions

Traditional solution labels use w/w, w/v, and v/v for three different concentration forms. State the numerator and the final quantity every time, because the same percentage means a different amount in each form.^{1,21 Thompson A, et al. Guide for the Use of the International System of Units (SI). 2008.2 Bishop ML, et al. Clinical Chemistry. 2023.}

The w/v form is dimensionally a mass concentration. SI writing therefore favors an explicit unit such as g/L. Reagent instructions may retain the traditional % w/v notation.

A 3.2% w/v NaCl solution has 3.2 g per 100 mL. For 250 mL, 3.2 × 2.5 = 8.0 g NaCl.

To convert a percentage to amount concentration, first express the mass concentration per liter. A 0.850% w/v NaCl solution contains 8.50 g/L. Using a molar mass of 58.44 g/mol gives 8.50 ÷ 58.44 = 0.145 mol/L. A 5.00% w/v H₂SO₄ solution contains 50.0 g/L. Dividing by the reaction-specific equivalent mass of 49.04 g/eq gives 1.02 N for complete neutralization of both protons.

Hydrated salts and concentrated stocks

The formula of a hydrate includes a fixed number of water molecules. Use the ratio of molar masses to adjust the required mass:

Mass of hydrate = required anhydrous mass × (molar mass of hydrate ÷ molar mass of anhydrous salt)

To replace 12.0 g of anhydrous MgSO₄ with MgSO₄·7H₂O, calculate 12.0 × (246.5 ÷ 120.36) = 24.6 g MgSO₄·7H₂O.

Concentrated-reagent calculations use the reagent assay as a mass fraction and the solution density. Specific gravity is a dimensionless relative-density ratio. Its numerical use depends on the stated reference temperature and water density.^{44 International Union of Pure and Applied… Relative density. 2025.} For an examination problem that gives HCl a specific gravity of 1.18, treats that value as 1.18 g solution/mL, and states 36.0% w/w HCl:

Unrounded HCl mass per milliliter = 1.18 g/mL × 0.360 = 0.4248 g/mL

Stock concentration = 424.8 g/L ÷ 36.46 g/mol = 11.65 mol/L, approximately 11.7 M

When stock and working concentrations use matching units, apply C1V1 = C2V2. To prepare 750.0 mL of 0.800 M HCl from this stock, retain the unrounded concentration:

V₁ = (0.800 mol/L × 750.0 mL) ÷ [(424.8 g/L) ÷ (36.46 g/mol)] = 51.5 mL

For an actual reagent, follow the current label, certificate, safety data sheet, and laboratory procedure. When diluting concentrated acid with water, add the acid slowly to the water while using the required ventilation and protective equipment.^{55 National Institute for Occupational Safety… Engineering Controls Database. 2018.}

Use molar mass to convert between mass concentration and amount concentration. To convert mg/dL to mg/L, multiply by 10. Then:

mmol/L = mg/L ÷ molar mass in mg/mmol

For a specified ion:

mEq/L = mmol/L × absolute ionic charge

A magnesium result of 2.10 mg/dL equals 21.0 mg/L. Using a molar mass of 24.305 mg/mmol gives 21.0 ÷ 24.305 = 0.864 mmol/L. Multiplying by the Mg²⁺ charge gives 1.73 mEq/L.

A glucose result of 108 mg/dL equals 1,080 mg/L. Using a molar mass of 180.16 mg/mmol gives 1,080 ÷ 180.16 = 5.99 mmol/L. Report patient results in the unit your laboratory has validated, using its own conversion factor.

An ideal blank-corrected solution with one absorbing species follows the Beer-Lambert relationship:^{66 International Union of Pure and Applied… Beer–Lambert law. 2025.}

A = εbc

In this equation, A is absorbance, ε is the molar absorption coefficient, b is path length, and c is the amount concentration. Absorbance is dimensionless. With path length in centimeters and concentration in mol/L, ε is commonly expressed in L·mol⁻¹·cm⁻¹. Its coherent SI unit is m²/mol.

A standard curve plots signals from known calibrators and applies the accepted calibration model to determine an unknown concentration. An ideal blank-corrected Beer-Lambert line passes through the origin. An empirical calibration may have an intercept or use a validated nonlinear model. The measurement procedure defines the calibration model and interval.^{77 Clinical and Laboratory Standards Institute. Evaluation of Linearity of Quantitative Measurement Procedures. 2020.}

Linearity studies use materials that span the claimed interval and concentrations important for clinical decisions. The study design, allowable deviation from linearity, and calibration model come from the measurement procedure and its validation plan. A single concentration provides evidence only at that concentration.

A validated linear calibration has an intercept of 0.005 and a slope of 0.00250 dL/mg:

A = (0.00250 dL/mg)C + 0.005

For an unknown with A = 0.305, the concentration is:

C = (0.305 − 0.005) ÷ 0.00250 dL/mg = 1.20 × 10² mg/dL

Instrument bandpass, stray light, high concentration, turbidity, and shifts in chemical equilibrium can cause the response to depart from the ideal relationship. Many absorbance systems lose useful relative precision as absorbance approaches about 2. The method's established interval sets the actual limit. Interpolate only within the validated calibration and measuring intervals. A result beyond the validated measuring interval requires the procedure's dilution and reanalysis or other approved follow-up. The highest calibrator concentration and the verified upper measuring limit are established separately, and they need not coincide. If no validated dilution is available, use the procedure's supported limit-form report, such as > upper limit.^{3,73 Clinical and Laboratory Standards Institute. Establishing and Verifying an Extended Measuring Interval… 2018.7 Clinical and Laboratory Standards Institute. Evaluation of Linearity of Quantitative Measurement Procedures. 2020.} The Core Chemistry Calculations lesson covers absorbance and percent-transmittance arithmetic. Quantitative Instrument Measurement Principles covers optical components and instrument-specific sources of nonlinearity.

Mean, median, and mode are different measures of a data set's location.^{88 National Institute of Standards and… Measures of location.}

Consider six chloride results: 107, 102, 110, 105, 108, and 104 mEq/L. In order, they are 102, 104, 105, 107, 108, and 110. Their mean is 636 ÷ 6 = 106 mEq/L, their median is (105 + 107) ÷ 2 = 106 mEq/L, and the set has no mode. In an ideal normal population, mean, median, and mode coincide. Similar sample values may be consistent with symmetry. Assess distribution shape separately with a histogram, normal probability plot, or another suitable method.

Control limits describe individual control results under a defined statistical-control procedure. When a hematocrit control range of 34% to 46% is explicitly defined as mean ±2 standard deviations (SD), its midpoint is 40%. The 12-percentage-point span equals 4 SD, so 1 SD is 3 percentage points. The ±1 SD interval is 37% to 43%, and the ±3 SD interval is 31% to 49%. This reverse calculation applies only because the stated limits are symmetric and explicitly defined from the mean and SD. Quality Control, Method Evaluation, and Quality Management covers control-chart assumptions and interpretation.

A confidence interval quantifies sampling uncertainty in an estimated population parameter. For independent observations from an approximately normal population with unknown population SD, a two-sided 95% interval for the mean uses:^{99 National Institute of Standards and… Confidence limits for the mean.}

95% confidence interval for the mean = sample mean ± t_0.975,n−1 × (sample SD ÷ √n)

Treating the six chloride results as an independent random sample, the sample SD is 2.8983 mEq/L and the standard error is 2.8983 ÷ √6 = 1.1832 mEq/L. The two-sided 95% t critical value with 5 degrees of freedom is 2.571:

106 ± (2.571 × 1.1832) = 106 ± 3.042, or 103.0 to 109.0 mEq/L

Over many repeated samples collected under the same conditions, about 95% of intervals calculated this way would contain the population mean. Larger samples narrow the interval when variability remains similar. A narrow interval cannot correct biased sampling or dependent observations.

Analytical performance and diagnostic performance are separate characteristics. Clinical laboratory literature often uses analytical sensitivity for low-concentration detection capability and analytical specificity for response to the intended measurand in the presence of potentially interfering substances. Diagnostic sensitivity and diagnostic specificity compare binary classifications with a designated reference standard for a defined target condition, threshold, population, specimen, and setting. The reference standard establishes target-condition status independently of the candidate test result.^{2,10,112 Bishop ML, et al. Clinical Chemistry. 2023.10 US Food and Drug Administration. Statistical Guidance on Reporting Results from Studies… 2007.11 Clinical and Laboratory Standards Institute. Evaluation of Qualitative, Binary Output Examination Performance. 2023.}

Metrological terminology separates these concepts further. IUPAC defines measuring-system sensitivity as the change in indication divided by the corresponding change in the measured quantity, commonly represented by calibration slope. IUPAC uses analytical selectivity for discrimination of the measurand from other substances.^{12,1312 International Union of Pure and Applied… Sensitivity of a measuring system. 2025.13 International Union of Pure and Applied… Analytical selectivity. 2025.} A calibration slope, limit of detection, interference response, and diagnostic sensitivity therefore require separate names and evaluations.

Sensitivity = TP ÷ (TP + FN) × 100%

Specificity = TN ÷ (TN + FP) × 100%

Positive predictive value (PPV) = TP ÷ (TP + FP) × 100%

Negative predictive value (NPV) = TN ÷ (TN + FN) × 100%

A hypothetical high-sensitivity cardiac troponin rule is evaluated for acute myocardial infarction in 500 emergency-department patients against a designated reference standard. The results are TP = 150, FP = 10, FN = 8, and TN = 332. The two-sided 95% Wilson score intervals quantify binomial sampling uncertainty for each fraction. Selection and reference-classification bias require separate study-design assessment.^{10,1110 US Food and Drug Administration. Statistical Guidance on Reporting Results from Studies… 2007.11 Clinical and Laboratory Standards Institute. Evaluation of Qualitative, Binary Output Examination Performance. 2023.}

Sensitivity and specificity estimated in one study may differ in another setting because they depend on the threshold, reference classification, range of participants, and study design. PPV and NPV also vary with prevalence and with how the study population was assembled. If the hypothetical participants form a representative cohort, the predictive values describe that cohort at its 31.6% observed prevalence. Predictive values from an enriched case-control study cannot be transferred directly to a clinical population. The Preanalytic, Analytic, and Postanalytic Quality Assessment lesson applies sensitivity, specificity, and pretest probability through Bayes reasoning.

Equivocal results and incomplete reference testing

An equivocal result remains an observed outcome after the assay's prescribed repeats. Account for it by reference status and explain its handling in the analysis. Deleting unresolved results can make performance appear better by removing difficult specimens. A binary table alone does not describe a test that retains a third category.^{1010 US Food and Drug Administration. Statistical Guidance on Reporting Results from Studies… 2007.}

Partial verification bias occurs when reference testing depends on the candidate result. Verifying only positive candidate results leaves false negatives unknown, so the ordinary sensitivity calculation lacks its full denominator. A planned subset design needs suitable sampling and an adjusted statistical analysis. These are FDA reporting recommendations and are not legal requirements.^{1010 US Food and Drug Administration. Statistical Guidance on Reporting Results from Studies… 2007.}

Combined-test denominators

For a rule requiring both assays to be positive, the algorithm-negative group contains three patterns: positive/negative, negative/positive, and negative/negative. Sum the reference-positive counts across all three for the algorithm's false negatives. Sum the reference-negative counts for its true negatives. The residual condition percentage is FN ÷ (FN + TN) × 100, the complement of NPV. Use paired observations. Multiplying separate assay probabilities is valid only with evidence for conditional independence within each reference-status group.^{1414 Gardner IA, et al. Conditional dependence between tests affects the diagnosis and… 2000.}

Comparison with a non-reference method

When a candidate test is compared with a routine method that is not a reference standard, the cells record agreement with the comparator. They do not establish the target condition.^{1010 US Food and Drug Administration. Statistical Guidance on Reporting Results from Studies… 2007.}

Positive percent agreement (PPA) = a ÷ (a + c) × 100%

Negative percent agreement (NPA) = d ÷ (b + d) × 100%

State the direction, because reversing the candidate and comparator changes the denominators. FDA's example with a = 40, b = 5, c = 4, and d = 171 gives PPA 40/44 = 90.9% (95% Wilson score interval, 78.8% to 96.4%) and NPA 171/176 = 97.2% (95% Wilson score interval, 93.5% to 98.8%). Overall agreement is 211/220 = 95.9%. Overall agreement or kappa alone can hide poor agreement in one result category and can change with the frequency of the condition. Report the positive and negative agreement pair with their fractions and intervals.