Week 2: Spectrophotometry Spectrophotometry Beer’s law Standard curve Dilution problems Quality control Control samples Levey-Jenning chart Shift and trend Out of control
Spectrophotometry Darkness = light absorption Corresponds to concentration A = 2 - log(%T) Light [Solution] Transmission
Beer’s Law A = abc = kc A = absorbance a = absorptivity b = light path length c = concentration k = constant (k-value)
Absorption Measurement Colorimeter with filter Spectrophotometer Prism Diffraction grating
Spectrophotometer Light source: tungsten Monochromator: diffraction grating Cuvette Photodetector: PMT, photocell Readout
Colman Spectrophotometer
Spectrophotometry 1. Turn instrument on 2. Select correct wavelength 3. Block light, set Zero (no light = infinity absorption = 0% T) 4. Choose and clean cuvette 5. Open light, insert Blank (maximum light = no absorption = 100% T) 6. Measure absorption of Standards, Controls and Patient samples to 3rd decimal place
Standards Precisely prepared = known concentration Usually pure solution of single compound Plot absorbance vs concentration: standard curve
Beer’s Law If standard curve is linear, use Beer’s law k = A std /C std = A unk /C unk C unk = A unk /k
For example… Std 1 = 100 mg% glucose Std 2 = 200 mg% glucose Abs Std 1 = Abs Std 2 = Abs Pt 1 = 0.200
Therefore… k Std 1 = 0.400/100 = k Std 2 = 0.800/200 = Average k =
Then… Since Abs Pt 1 = And C = A/k C of Pt 1 = ÷ = 50 mg% Since glucose reference range is mg%, she is hypoglycemic!
Units dL = 100 mL %(w/v) = g/dL mg% = mg/dL %(v/v) = mL/dL CLIA 88: two or more standards are required for most tests; make sure k-values match
Murphys’s Law If something can go wrong, it will! Sources of errors: random vs technical Sample - pre-analytical Reagent Method Technique Reporting - post-analytical
How to Ensure Accuracy? Repeat tests many times (how many?) and take average Run another sample that was tested before along with patient samples and make sure its result is close to what it should be
Control Samples Similar in composition to patient sample Usually pooled from many donors Tested at least 30 times to calculate the average (target value) and allowable range of variation
Are You in Control? Was your control value close to the target value? Predetermined mean value How close is close enough? Within ± 2 standard deviation from the mean
Formulae Mean = Sum of all values = ∑ xi Sample population n ± 2 Std = ± 2 ∑ (xi - mean) 2 n - 1 √
Calculate the QC Statistics #Chol (mg%) Mean = 500mg% ÷ 5 = 100 mg%
Calculate the QC Statistics #Chol (mg%)d d = x i - mean
Calculate the QC Statistics #Chol (mg%)dd ∑ d 2 = 38
Calculate the QC Statistics #Chol (mg%)dd ± 2 Std = ± 2 ∑ (xi - mean) 2 n - 1 √ = ± √ = ± 6.2 mg%
That Means… Target value = 100 mg% Allowable range = ± 6.2 mg% Acceptable control range: mg% If your control value was within the above range, you are in control.
Levey-Jennings QC Chart Graphical record of QC Better able to spot shift and trend Visualize degree of random distribution CLIA 88 requires at least two levels of controls per test
Shift Trend
Westgard’s Rule 1 2S : warning 1 3S : usually random; re-run 2 2S : both of consecutive violation R 4S : usually random; large range 4 1S : indication of shift 10 x : shift If all above are OK, then accept result
Precision vs Accuracy Reproducibility Close to the true value Estimation of true value? In clinical labs, we need both plus speed: efficiency