Selected Phenolic Compounds in Mainstream Cigarette Smoke, CORESTA Collaborative Study and Recommended Method *

The laboratories that conducted the testing along with the product volunteers are listed in alphabetical order in Table 1. To ensure anonymity of the results, each laboratory was given a unique study code for data exchange and reporting.

Participating laboratories were asked to follow a supplied study protocol, a draft of a CORESTA Recommended Method (CRM), and Health Canada T-114 (14) analytical method for the determination of phenols. Samples analyzed for the study are listed in Table 2 with a range of blend and cigarette designs that represent the market and a wide range of total particulate matter (TPM) and analyte yields. In accordance with the study protocol, participating laboratories were asked to obtain reference cigarettes and monitor test pieces directly from suppliers (19–20). Other samples were commercial products from various regions provided by the study volunteers listed in Table 1. Five replicates of each product with each smoking regime were requested to be generated. Reportable measures included puff count, TPM, hydroquinone, resorcinol, catechol, phenol, o-cresol, m-cresol, and p-cresol.

The yields of phenols collected under ISO 3308 (non-intense) and ISO 20778 (intense) smoking regimes were reported in units of microgram per cigarette (µg/cig) in the templates provided to participants.

Following the recommended method, conditioned cigarettes were smoked using an automated 20-port constant volume smoking machine, onto a conditioned, pre-weighed glass fiber filter disc (pad).

TPM was determined gravimetrically and puff counts were as recorded. The pads were extracted using 40 mL of 1% aqueous acetic acid with 30 min of agitation using a wrist-action shaker. Extracts were diluted using 1% acetic acid based on total TPM. Samples below 15 mg TPM were tested undiluted. Samples with 15–60 mg TPM were diluted 2:5 and samples with 60–100 mg TPM were diluted 1:5 prior to analysis.

Samples were analyzed within 24 hours of collection and extraction as hydroquinone is subject to rapid oxidation. Samples were analyzed using high performance liquid chromatography with selective fluorescence detection (HPLC-FLD).

Detector settings are included in Table 3.

The mobile phase consisted of: 1% acetic acid (HOAC) (aqueous; solvent A); 1% HOAC (in methanol; solvent B) with a flow rate of 0.8 mL/min and an elution gradient as noted in Table 4.

Sample injection volume was 10 µL to 20 µL onto a pentafluorophenylpropyl analytical column (4.6 mm × 150 mm × 3 µm).

Data consistency was verified using graphical and numerical outlier detection techniques. Mandel’s h plot and Grubbs test were used to evaluate between-laboratories consistency.

Mandel’s k plot and Cochran’s test were used to evaluate within-laboratory consistency. Grubbs and Cochran tests are less likely to qualify data points as outliers than their corresponding graphical methods and to avoid excessive data exclusion. Thus, the final decision on straggling (0.95) and outlying (0.99) data was made using Cochran’s and single-iteration Grubbs tests.

Straggling data were retained in the dataset but outlying data were removed for descriptive statistics reporting. Overall mean, standard deviation, repeatability (r), and reproducibility (R) values were determined according to ISO 5725-216 (18).

The study design allowed for evaluation of the robustness of the method across the linear range for analytes of interest and confirmed the method’s effectiveness with a variety of products.

Thus, the study samples chosen represented a wide variety of TPM yields, tobacco blends, and product construction parameters including varied filter ventilation level and inclusion of a filter additive for one study sample. Non-intense and intense smoking regimes were used to verify applicability with the recommended method.

Smoking related measures, including puff count and TPM, had low variability for both smoking regimes and across all products. Sample 7 (activated charcoal filter and high filter ventilation) showed the highest variability among the products tested. Additionally, no significant differences in results were noted based on choice of smoking machine type (i.e., linear or rotary). The method appears to distinguish samples appropriately. As expected of particulate phase and pad-trapped semi-volatile constituents, the analyte yields were directly correlated to TPM yields. When normalized to TPM, phenols showed predictable trending based on blend type. For example, Virginia (i.e., fluecured) products had higher phenolic:TPM values compared to American blend samples containing Burley tobacco.

In general, CM7 had the highest phenolic:TPM values. This is consistent with previous research indicating that flue-cured tobacco may generate higher levels of some phenolic HPHCs (10, 21).

Several data sets included non-numeric (i.e., < LOQ) reporting for resorcinol, phenol, and the cresol compounds. These values were excluded from outlier evaluation and calculation of descriptive statistics and repeatability (r), and reproducibility (R) values.

Results of graphical outlier detection methods (Mandel’s h and k plots) are not displayed in this report. During the first iteration, there were 10 stragglers and 8 outliers for the non-intense regime, and 13 stragglers and 7 outliers for the intense regime. A second iteration of the test did not reveal any further outliers. Grubbs single outlier test was employed on the lab means and identified the following outliers and stragglers. There were 4 stragglers and 8 outliers in the non-intense regime, and 10 stragglers and 13 outliers for the intense regime. All outliers were removed from the dataset. All straggling data was retained.

Mean, r, and R results with outliers removed are displayed in Table 5 and Table 6. Values are typically above limits established in ISO 8243 (22) for “tar”, nicotine, and carbon monoxide (TNCO). This is not unexpected given the much lower levels for phenols (µg) than TNCO measures (mg). Additionally, the highest values for r and R noted in the study were for results near LOQ at limits of the method’s capabilities.