Statistical Tools for SWGTOX Method Validation of 11 Benzodiazepines in Whole Blood by SPE and GC/MS Szabolcs Sofalvi, M.S., D-ABFT-FT Cleveland, Ohio
Disclaimer Neither I nor any member of my immediate family has a financial relationship with a company as defined in the AACC policy on disclosure of potential bias or conflict of interest. SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 2
Introduction Driving force for a new method? SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 3
Chromatographic Quality SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 4
Objective Validate Benzodiazepines (11) Assay 5 batches (different days/analysts) Customized Microsoft Excel Template ANOVA (8) Calibration Model (5) LOQ (1) Bias and Precision (2) [Low/High QC] LODs (Estimated) SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 5
Method Sample Size (2mL) Deuterated Internal Standards Solid Phase Extraction UCT Clean Screen (ZSDAU020) Derivatized with MTBSTFA Agilent 7890A/5975C GC/MS Phenomenex ZB-50 GUARDIAN Part No. (7CD-G004-08- GGT) SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 6
Results Analyte Lorazepam α-hydroxymidazolam Clonazepam Midazolam Alprazolam α-hydroxyalprazolam Temazepam Diazepam Oxazepam Nordiazepam Linearity [ng/ml] 6-100 10-400 20-800 SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 7
Response Ratio 1.8000 1.6000 1.4000 1.2000 1.0000 0.8000 0.6000 0.4000 0.2000 0.0000 Calibration Model Alprazolam 0 100 200 300 400 Concentration (ng/ml) Trial 1 Trial 2 Trial 3 Trial 4 Trial 5 Average SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 8
Drug: Alprazolam Calibration Model Trial 1 SUMMARY OUTPUT Regression Statistics Multiple R 0.99988734 R Square 0.99977469 Adjusted R Square 0.9997296 Standard Error 0.008928472 1. Regression Statistics Observations 7 ANOVA df SS MS F Significance F Regression 1 1.76867 1.76867037 22186.7 2.5874E-10 Residual 5 0.000399 0.0000797 Total 6 1.7691 Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Intercept 0.064058099 0.005057 12.6680772 5.4E-05 0.0510596 0.0770566 X Variable 1 0.003618896 2.43E-05 148.951985 2.6E-10 0.0639956 0.0036814 2. ANOVA RESIDUAL OUTPUT Observation Predicted Y Residuals Standard Residuals 1 0.100247064 0.01040 1.275524 2 0.15453051 0.00789 0.968596 3 0.245002921-0.00982-1.204992 4 0.425947743-0.00846-1.037939 5 0.787837388-0.00418-0.512634 6 1.149727032-0.00204-0.250220 3. Residual Output 7 1.511616677 0.00621 0.761665 SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 9
Calibration Model 1. Regression Statistics Concentration Response Ratio Regression Statistics Multiple R 0.99988734 =CORREL($B$7:$B$13,F7:F13) select ranges for x,y (use $ sign) R Square 0.99977469 =CORREL($B$7:$B$13,F7:F13)^2 typed rsquare1 in upper left corner Adjusted R Square 0.9997296 =1-((1-rsquare1)*(n-1)/(n-2)) Residuals Standard Error 0.008928472 =SQRT(SUMSQ(D7:D13)/(n-2)) typed serror1 in variable name cell Observations 7 =COUNT($B$7:$B$13) entered below L7 (ULOL) cell & type n in upper left corner Look for this entry on the next slide! SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 10
Calibration Model 1. Regression Statistics SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 11
Residuals 101 y [Response Ratio] y-ŷ ŷ =m*x + b x [Concentration] SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 12
Calibration Model 2. ANOVA ANOVA df SS MS F Significance F Regression 1 1.76867 1.76867037 22186.7 2.5874E-10 Residual 5 0.000399 0.0000797 Total 6 1.7691 Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Intercept 0.064058099 0.005057 12.6680772 5.4E-05 0.0510596 0.0770566 X Variable 1 0.003618896 2.43E-05 148.951985 2.6E-10 0.0639956 0.0036814 SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 13
Calibration Model 3. Residual Output RESIDUAL OUTPUT ŷ N74 Observation Predicted Y Residuals Standard Residuals 1 0.100247064 0.01040 1.275524 =(N74)/STDEV(D$7:D$13) 2 0.15453051 0.00789 0.968596 3 0.245002921-0.00982-1.204992 4 0.425947743-0.00846-1.037939 5 0.787837388-0.00418-0.512634 6 1.149727032-0.00204-0.250220 7 1.511616677 0.00621 0.761665 SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 14
Standardized Residual Plot Alprazolam 3.000000 Standardized Residuals 2.000000 1.000000 0.000000-1.000000-2.000000-3.000000-20 80 180 280 380 Concentration (ng/ml) Trial 1 Trial 2 Trial 3 Trial 4 Trial 5 Average SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 15
Limit of Quantitation (LOQ) L1 Concentration 10 ng/ml Trial 1 Trial 2 Trial 3 Trial 4 Trial 5 Calculated Mean Replicate 1 12.87 12.80 10.27 11.038 Replicate 2 10.33 11.52 10.42 Bias Replicate 3 10.89 11.21 9.03 10.38% SUMMARY Groups Count Sum Average Variance SS Trial 1 3 34.09 11.36 1.780933 3.561867 Trial 2 3 35.53 11.84 0.710433 1.420867 Trial 3 3 29.72 9.91 0.58203 1.164067 SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 16
Limit of Quantitation (LOQ) ANOVA Source of Variation SS df MS F P-value F crit Between Groups 6.102956 2 3.051478 2.978601 0.126347 5.143253 Within Groups 6.1468 6 1.024467 Total 12.24976 8 Within Run CV (%) 9.17% Between Run CV (%) 11.81% SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 17
LOQ = 10 ng/ml SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 18
Bias and Precision (QCs) Low QC Concentration 125 ng/ml Trial 1 Trial 2 Trial 3 Trial 4 Trial 5 Calculated Mean Replicate 1 125 123 125 126 124 126 Replicate 2 130 125 125 131 122 Bias Replicate 3 128 126 130 134 123 1.17% High QC Concentration 250 ng/ml Trial 1 Trial 2 Trial 3 Trial 4 Trial 5 Calculated Mean Replicate 1 252 257 262 265 262 261 Replicate 2 254 255 271 265 252 Bias Replicate 3 255 274 264 268 254 4.27% SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 19
SUMMARY Bias and Precision(ANOVA) Groups Count Sum Average Variance SS Trial 1 3 383 127.667 6.3333 12.66667 Trial 2 3 374 125 2 4.666667 Trial 3 3 380 126.667 8.3333 16.66667 Trial 4 3 391 130 16 32.66667 Trial 5 3 369 123.000 1.0000 2 ANOVA Source of Variation SS df MS F P-value F crit Between Groups 95.06667 4 23.76667 3.461165 0.050635 3.47805 Within Groups 68.66667 10 6.866667 Total 163.7333 14 Within Run CV (%) 2.07% Between Run CV (%) 2.80% SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 20
Estimated LOD from Calibration Curves Limit of Detection (LOD) Estimated Slope y-intercept Trial 1 0.0036 0.0641 Trial 2 0.0037 0.0691 Trial 3 0.0037 0.0698 Trial 4 0.0036 0.0580 Trial 5 0.0038 0.0608 Average 0.0037 0.0643 Standard Deviation 0.0001 0.0051 =SLOPE('Calibration Model'!G7:G13,'Calibration Model'!B7:B13) =INTERCEPT('Calibration Model'!G7:G13,'Calibration Model'!B7:B13) Estimated LOD 4.56 ng/ml =(3.3)*(G14)/D13 LLLLLL = 33. 33 SSSS yy iiiiiiiiiiiiiiiiii AAAAAA SSSSSSSSSS SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 21
Conclusions Improved chromatographic quality: Ion ratios Peak shapes Number of data points/peak Tailing Factor (< 2.00) SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 22
Conclusions Customized Microsoft Excel: Instant results Accurate No need to reformat every time Validation Summary Report Advantage: Other validations! (e.g. Speaker 42) SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 23
Harold E. Schueler Xin Xu (Speaker S42) Claire K. Kaspar Cassandra L. Clyde Eric S. Lavins Thomas P. Gilson Acknowledgments SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 24
Questions Email: ssofalvi@cuyahogacounty.us SOFT 2016 Dallas, TX Development of Statistical Tools for Method Validation 25