Tobacco Documents Online

February 1b, 1yytS REVIEW OF Development and validation of sensitive method for determination of serum cotinine in smokers and nonsmokers by liquid chromatography/atmospheric pressure ionization tandem mass spectrometry Bemert Jr., J. T., Turner, W. E., Pirkie, J. L., Sosnoff C. S., Akins, J. R., Waldrep, M. K., Ann, Q., Covey, T. R., Whitfield, W. E., Gunter, E. W., Miller, B. B., Patterson Jr.,`D. G., Needham, L., L., Hannon, W. H., Sampson, E. J. Clin Chem 43(12) 2281-2291 (1997) Author's Abstract We describe a sensitive and specific method for measuring cotinine in serum by HPLC coupled to an atmospheric pressure chemical ionization tandem mass spectrometer. This method can analyze 100 samples/day on a routine basis, and its limit of detection of 50 ng/L makes it applicable to the analysis of samples from nonsmokers potentially exposed to environmental tobacco smoke. Analytical accuracy has been demonstrated from the analysis of NIST cotinine standards and from comparative analyses by both the current method and gas chromatography/high-resolution mass spectrometry. Precision has been examined through the repetitive analysis of a series of bench and blind QC materials. This method has been applied to the analysis of cotinine in serum samples collected as part of the Third National Health and Nutrition Examination Survey (NHANES III). Review Summary Overall, this is an excellent paper. The analytical approach, validation and QC methods are exceptionally well done. Liquid Chromatography-Atmospheric Pressure Ionization Tandem Mass Spectrometry (LC-API/MS/MS) is an immensely powerful analytical tool, and is one of the fastest growing tools for organic quantitative measurements at low concentration. In the detailed review, a few minor points will be mentioned that perhaps could have improved the work. During a 4-year period, >32,000 serum samples have been analyzed using a single instrument, which is testament to the reliability and ruggedness oft`ne method and the associated hardware. The single most outstanding weakness in the paper is that cotinine taken alone and not in conjunction with other nicotine metabolites is not a good marker for nicotine exposure because of the interpersonal differences in metabolism rates and metabolite distribution. This paper reports the development and validation of the LC-API-MS/MS method that was used in the Third National Health and Nutrition Examination Survey (NHANES 111) Philip Morris Europe Confidential REVIEW_LC APl-MS.DOC 1

reoruary 'io, 'iaass I study.1 With reference to the review by Benowitz,2 the authors conclude that cotinine is the biomarker of choice for determining exposure to tobacco by active smoking or by passive exposure. The NHANES III study was conducted well before the Benowitz review appeared. The usual points to support that position are given and need not be discussed here. The method was developed to provide sensitivity and selectivity for a high sample throughput for determinations of cotinine in serum. Comments will be made according to contiguous sections in the paper. Materials and Methods: Standards and reagents: The authors employed reasonable sources of standards and reagents. There is no reason to suspect other than quality materials were used. Calibration standards: If taken as stated, only a single preparation of a set of cotinine calibration standards was made for the entire NHANES study. This nieans that all of the data are dependent upon that preparation. Again, there is nothing to suggest an error, but if one were made in the preparation of that standard, then all results would be in error. This is risky analytical practice. Admittedly, the authors do not state whether the set of 12 standards was prepared by dilution of a single stock solution, or if each of the 12 was prepared separately, nor do they state what is mean by analysis of the standards to confirm their suitability. QC materials: Reasonable QC precautions were taken. However, it is less important that the QC sample insertion protocol was "computer generated" than to describe the algorithm. At least the authors used QC samples which is frequently not the case, or not mentioned in publications. Preparation of materials: In the case of previous analytical methods, the very low levels of cotinine that are present in tobacco and tobacco smoke were not of significance as a potential source of contamination. However, at the level of detection of the method in this report it is a potential issue. Using this as a basis, smoking was prohibited in the entire building for more that one year prior conducting cotirEine determinations, and anyone involved in sample handling was Philip Morris Europe Confidential REVIEW_LC AP!-MS.DOC 2

reoruary -iti, 7yy25 required to be a non-smoker. Sample containers predesignated for use with high and low cotinine concentrations were kept separate at all times. lnstrumentation: A Perkin Eimer Sciex API III atmospheric pressure ionization triple quadrupole mass spectrometer was used for this work. The HPLC system used a deactivated C18 column in an Hewlett Packard Model 1090L equipped with autosampler and Sciex software to control sample injection and data analysis. Sample Preparation: Because of the large range of values of serum cotinine that can be found when smokers and non-smokers are included, the samples and QC samples were screened with an enzyme-linked immunoassay (ElA) method. Based on the EIA resuft, each sample was placed in a "high" or "low" category for LC/MS/MS determinations. Extraction and Cleanup; Sample Analysis; Standards Analysis; Data Analysis'and Additional Analyses: These sections present detailed descriptions of the procedures followed. These details appear to present a sound scheme for the analytical measurements and data treatment. Results and Discussion: Analytical Scheme: As described above, EIA screening using a cutoff of 25µg/L was used to separate samples into two groups. The high group was diluted so that the highest value would not exceed 25µg/L. Calibration Curve: The calibration curve (Fig. 2) was prepared using 302 replicated runs of the standard sets (a total of 3624 runs). This is a remarkable number of replicates. However, as mentioned above the statement that a single set of standards was prepared makes this curve and excellent test of precision, but provides no information as to its accuracy. A major advantage of the method is its low level of detection. (The authors appear to confuse this with sensitivity, which is the slope of the calibration curve or change in response per urrit change in concentration.) The method of calculation of the Philip Morris Europe Confidential REVIEW-LC API-MS.DOC 3

reuruary 10, iV-Vo lower limit of detection (LOD) is an optimistic one. Considering only the instrumental noise and LOD of about 26 ng/L was estimated by estimating the concentration that yields a signal equal to 3 times the standard deviation in the limiting noise. Recognizing that this was quite optimistic, the authors chose to use the mean standard deviation of a water blank and obtained and LOD of approximately 50 ng/L. Even this does not include the noise resulting from sampling and sample treatment. Furthermore, the important parameter for a quantitative method is the level of quantification (LOQ), which is not mentioned by the authors. This value should be 2-3 times the LOD, or in this case 100-150 ng/L. The impact of these statistical choices is in part reflected in Fig. 4. Above about 100 ng/L, the ratio of the confirmation to the quantitation ion has relatively constant precision, but the precision rapidly decreases below that concentration. !on chromatograms: The authors describe the ion area ratios that are used and these are appropriate. Recoveries: Recoveries were estimated in two ways and averaged about 70% except for samples of the lowest concentration which gave about 60% recovery. The authors have evidence that some loss occurs in the protein pellet in the step in which trichioroacetic acid is added. The reduced problems with emulsion formation later in the sample work up were thought to be justification for the step. Accuracy Evaluations: Comments made earlier concerning reliance on a single set of standards are to some degree tempered by accuracy evaluations. A series of aqueous standards prepared analytically from cotinine perchlorate were analyzed, as were aliquots of NIST Reference Material 8444. Agreement was acceptable. Supplemented Serum Analyses: Serum samples spiked with cotinine were analyzed blind by an analyst not associated with the project. These gave nominal means and precision. Comparisons with GC/IAS: Split samples were analyzed by LC-API/MS/MS and C.C/MS. The results compared exceptionally well. This step was an important part of the overaif vaiidation of the LC-APf/MS/MS method. However, it is likely not W Philip Morris Europe Confidential REVIEw_LC API-MS.DOC 4

I cuS ua4 y iv, a u.7u "necessary" to use high resolution GC/MS with deuterated cotinine as an internal standard as stated by the authors. It is probable that GC/MS/MS would have achieved a similar result. Interferences: Serotonin and ascorbic acid both have masses of 176 and are present in blood. Tests with standards showed that these compounds do not interfere with cotinine. The only potential interference known at publication is Pemoline, a CNS stimulant. Thus, use of the confirmation ratio is important. Cotinine Stability Consistent with the rest of the paper, the authors made a point to check the stability of cotinine in high and low serum pools. There were no detectable changes in serum stored at room temperature, or at 37 °C for up to 6 weeks. The authors also reported no detectable changes in cotinine concentration in urine samples stored at -60 °C for more than 4 years. Citations 1. Pirkle JL, Flegal KM, Bemert JT, Brody DJ, Etzel RA, Maurer KR., Exposure of*the US population to environmental tobacco smoke. JAMA 1996;275(16):1233-1240. 2. Benowitz NL., Cotinine as a biomarker of environmental tobacco smoke exposure. Epidemiol Rev 1996; 18(2):188-204. Philip Morris Europe Confidential REVIEW_LC API-MS.DOC 5