Comparison of Select Analytes from Different Laboratories in Tobacco and Smoke for Commercial Cigars Across a Range of Designs

The study design was a randomized two-factor arrangement. The two major factors were: Products (with six levels defined by size or weight, making technique and tipping) and Laboratory (with three levels as Lab 1, 2, 3). Product selection was based on inclusion of typical product types with standard and challenging design parameters (e.g., cylindrical vs. non-cylindrical designs). Laboratory selection criteria included capability and expertise to test for all product types and analytes and measures, and use of onscope for ISO 17025 methodology to conduct the study. Note, our original target was five laboratories. No other labs evaluated for the study met all criteria for inclusion.

Six conventional cigar products were purchased through wholesale or retail sources for testing. Information for each of the products is listed in Table 1 and the products are displayed in Figure 1. The physical measures noted were estimated based on testing a small aliquot of each product during the development of the study design.

Figure 1.

Cigar products tested. Samples are ordered from top to bottom in the photograph as Sample A, Sample B, Sample C, Sample D, Sample E, and Sample F.

ISO 17025-accredited laboratories used in-house analytical methodology to evaluate the cigar samples for selected HPHCs. Typically, methods used were adaptations of, or similar in principle to, existing cigarette or smokeless tobacco methods referenced herein.

No requirements for target method limits of quantitation (LOQ) or reporting conventions for data outside a method’s upper or lower limits were provided to participating labs. For samples with one or more < LOQ values, post-reporting calculations were based on the reported average.

Cigar weight, tobacco weight, cigar length, diameter, and resistance to draw (n = 20) were measured using conventional equipment in everyday use in industry laboratories with modifications as warranted by laboratories based on unique features of a given sample (12, 13).

Tobacco was isolated from other cigar components in preparation for analysis. As warranted, tobacco samples were oven-dried for consistent grinding; thus, analytes may be reported on an ‘as-is’ or a dry weight basis. Replicate analysis (n = 7) was conducted on separate-grind aliquots.

Nicotine: Ground tobacco was extracted using liquid/liquid extraction into organic solvent followed by gas chromatography with a downstream flame ionization detector (GC-FID) analysis (14) (Lab 2, Lab 3) or using aqueous solvent followed by continuous flow analysis (15) (Lab 1).

Ammonia: Ground tobacco was extracted using dilute acid. Filtered extracts were analyzed by ion chromatography (IC) (16).

Carbonyls: Acetaldehyde, crotonaldehyde, formaldehyde; tobacco aliquots were extracted and derivatized in a two-phase system with aqueous buffer and isohexane using 2,4-dinitrophenyl hydrazine (DNPH) as a derivatization agent. Extracts were subsequently analyzed by ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) (17).

Metals: Arsenic and cadmium; digested tobacco was analyzed by inductively coupled plasma mass spectroscopy (18).

Polyaromatic hydrocarbons (PAH): Benzo[a]pyrene (BaP); tobacco was extracted in methanol followed by solid phase extraction (SPE) clean-up and analysis by gas chromatography-mass spectrometry (GC-MS) in selected ion monitoring (SIM) mode (19).

Tobacco-specific nitrosamines (TSNAs): N-Nitrosonornicotine (NNN), 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK); ground tobacco was extracted using dilute ammonium acetate, and extracts were analyzed by liquid chromatography tandem-mass spectrometry (LC-MS/MS) (20).

Smoke collection and trapping: Cigar preparation and machine smoking requirements are described in CORESTA Recommended Method CRM 65 (11). As prescribed, one or two cigars were smoked per replicate (n = 7). Smoke was collected onto 44-mm, 55-mm, or 92-mm glass fiber filter pads with > 99% particulate trapping efficiency for each replicate analysis. For carbonyls, smoke was collected directly by two in-series impingers. For volatile organic compounds (VOC) determinations, single chilled impingers were placed in-line with the filter pads.

Method quality control measures were as-typical for each laboratory and included testing an in-house control chart sample or analyte-spiked matrix samples.

“Tar”, nicotine, carbon monoxide (TNCO): Total particulate matter (TPM) was determined gravimetrically. Filter pads were subsequently extracted with alcohol, extracts were analyzed by GC-FID for determination of nicotine and gas chromatography with a thermal conductivity detector (GC-TCD) for determination of water (21, 22). Nicotinefree dry particulate matter, “tar”, was calculated by subtracting nicotine and water values from the TPM value. Carbon monoxide was determined concurrently with smoke collection for nicotine and water using a tandem non-dispersive infrared (NDIR) analyzer (23).

Ammonia: Cigars were smoked through glass filter pads with in-line impingers containing dilute acid. Pads were extracted with the impinger solvent. Extracts were analyzed by ion chromatography (IC) (24).

Carbonyls: Acetaldehyde, acrolein, crotonaldehyde, formaldehyde; mainstream cigar smoke was trapped in impingers containing an acidified 2,4-dinitrophenylhydrazine (DNPH) solution. After smoking, trizma base was added to the combined solutions, which were subsequently analyzed by ultra performance liquid chromatography with a diode array detector (UPLC-DAD) (25).

Polyaromatic amines (PAA): 4-Aminobiphenyl, 1-aminonaphthalene, 2-aminonaphthalene; pads were extracted with dichloromethane followed solid phase extraction (SPE) clean-up and analysis by GC-MS in negative chemical ionization (NCI) mode (26).

Polyaromatic hydrocarbons (PAH): Benzo[a]pyrene (BaP); BaP analysis was conducted by extraction in methanol followed by SPE clean-up and analysis by GC-MS in - selected ion monitoring (SIM) mode (27).

Tobacco-specific nitrosamines (TSNAs): NNN, NNK; cigars were smoked onto glass filter pads which were subsequently extracted using dilute ammonium acetate. Extracts were analyzed by LC-MS/MS (28).

Volatile organic compounds (VOC): Acrylonitrile, benzene, 1,3-butadiene, isoprene, toluene; filter pads were extracted in methanol (MeOH), and extracts were analyzed by GC-MS in SIM mode (29).

The results obtained from these analyses were tabulated as mean ± one standard deviation for levels of selected compounds. As indicated, Analysis of Variance (ANOVA) was used to compare data among laboratories. Statistical analyses were performed using JMP 13.0.0 (SAS Institute, Inc. Cary, NC, USA). The significance level was established as p < 0.05 for all comparisons.

Data from each laboratory were compiled and evaluated to determine the range in mean among the sample set, the relative variability (% RSD, Relative Standard Deviation) among the sample types, and to generally compare the laboratories’ results typically through evaluation of relative range values. As a practical method for laboratory comparisons, two types of relative range were calculated and expressed as a percentage value and without regard for standard deviation or statistical significance as noted below for Lab Range (LR) and Lab Product Range (LPR). LR indicates the spread in reported results among all of the labs for a given product while LPR allows for a comparison among the labs for the spread in results across the product set. A low LR value and a similar LPR for results would indicate agreement in results among the laboratories. The LR value indicates how large a difference might we expect in reported values from one lab to another. The LPR value indicates how similar or different the reported values distinguish the samples from one another.

The value in determining physical parameters is that these measures may indicate the natural variability of product manufacture. Some combinations of tobacco weight, cigar length, and cigar diameter are control parameters during manufacturing for most products. Pressure drop may be controlled/measured and is anticipated to impact smoke analytes’ variability significantly.

Typically, the greater the variability in physical parameters, the greater the variability (% RSD and/or LPR) in smoking measurements and analytes. Smoke yields and variability correlate to tobacco weight; higher or more variable weight would likely result in more analyte or variability for an analyte, respectively. Pressure drop is a measure of relative resistance to airflow through the cigar column and is reported in units of mm of water. It is imperative to understand this parameter for machine smoking of cigars, given that airflow is established as a ‘constant’ under the smoking regime employed. Particulates’ yield is generally inversely proportional to pressure drop; higher airflow resistance would result in an analyte’s lower yield.

Each sample was tested for weight, length, diameter, and pressure drop (sometimes referred to as resistance to draw) (n = 20). Results collected for physical parameters from three different labs are displayed in Table 2. Data are presented as mean ± standard deviation with percent relative standard deviation (% RSD) as a parenthetical. Incidents of deviation included some cases where only one value was reported. For example, in the case of cigar weight, Lab 3 weighed all cigars together and divided by twenty. Additionally, Lab 3 did not report conditioned tobacco weight for samples A–D and only reported physical parameter data for one replicate of samples E and F. The range in nominal and reported values among the samples was quite broad, and the samples were easily distinguishable based on the significant differences in physical parameter measurements. For example, conditioned cigar weight among the samples ranged from less than 1.5 g/cigar to approximately 20 g/cigar. Diameter results ranged from less than 8 mm for Sample A to more than 20 mm for Sample F. Pressure drop results were inversely proportional to cigar length and diameter and ranged from approximately 270 mm H₂O for Sample A to 35 mm H₂O for Sample F.

Among the measured parameters in the study, physical parameters’ replicates showed relatively low variability compared to tobacco and smoke results. See percent relative standard deviation (% RSD) calculations in Table 2. Of note, the typical % RSD for CORESTA Monitor cigarette weights is less than 1% (30) compared to as high as 10% (Sample C) observed for cigar samples in this study. Minimal variability was observed for length and diameter. On the other hand, pressure drop % RSD values and laboratory range (LR) values are as high as 43% and 33%, respectively. Figure 2 displays the mean and three times standard deviation for the samples’ pressure drops. Sample A, with the highest pressure drop, has the lowest variability (~5%) and the lowest spread in reported results (10%). This is not surprising given that this sample type is the only design amenable to measurement with conventional testing equipment in use among the laboratories. Standardized equipment customized for cigar physical parameter testing is limited and cost-prohibitive. Thus, other cigars in the study present challenges to measuring pressure drop. While there is commercial, manual, or automated equipment available for cigarettes and cigars similar in diameter and length to cigarettes, larger cigars (Sample E and F), non-cylindrical cigars, and cigars with unusual shapes (Sample B) or mouthpieces (Sample D) must be measured with improvised, non-standardized equipment.

Figure 2.

Lab comparison: Pressure drop average of 7 replicates for each sample from three different labs. Error bars, if shown, are 3 × standard deviation.

Despite the challenges of measuring physical parameters for a diverse sample set, the three laboratories were able to generate mean data that was similar between the labs. Not surprisingly, the average conditioned cigar weight (g/cigar) values were consistent over the cigar sample set as demonstrated by the low LR values for each sample. In addition, the range of conditioned cigar weight means observed for each laboratory, LPR, spanned comparable ranges 234% to 245%. The other physical parameter means for Samples A–F in Table 2 also demonstrate similar consistency between the three laboratories. Overall, the difference in LPR values for the laboratories ranged from 2 to 12 indicating a similar level of discernability among the labs for the physical parameters; one would anticipate similar results from any of the three labs for products of this divergent range in design.

Traditional cigarette testing incorporates the use of monitor or reference cigarettes and smokeless tobacco products that serve as positive controls and provide quality metrics for standardized analytical methods. They are useful for comparing data, in context, across studies, or among different laboratories in the same study. Key examples are University of Kentucky reference cigarettes and CORESTA monitor cigarettes. Each of these reference products can serve as a single positive control and an indicator of method variability within and among laboratories for all analytes of interest. The manufacture, design, and function of these reference products are similar to those of commercial cigarettes. Only recently have finished cigar reference products been made available for testing. While there is an ongoing FDA-sponsored project to establish a set of certified cigar reference products (31), limited characterization data for standardized cigar products, particularly among different laboratories and smoke constituents, is available.

In the absence of well-established standardized references for cigars, each laboratory in this study was encouraged to use its internal measures to ensure experimental robustness. In particular, established internal acceptance criteria for test batches were employed.

Nicotine, ammonia, carbonyls, trace metals, BaP, and TSNAs (n = 7) were determined. Results shown in Table 3 are presented as dry weight basis mean ± standard deviation (% RSD) for before-grinding sampling. The spread in mean results was calculated and reported as LR in Table 3.

As with physical parameters measurements, the choice of samples included in the study afforded a wide range of measured values for LPR calculations. For example, nicotine ranged from approximately 8,000 µg/g to approximately 35,000 µg/g, and NNN showed a 21-fold range of results. BaP and crotonaldehyde were low (i.e., at or below LOQ) and indistinguishable among the samples supporting that these are not likely to be high valued measures for cigar tobacco characterization. The range in tobacco analyte levels is consistent with previous studies and is expected to be due to tobacco variety and growing and harvesting conditions (1).

In general, replicate analysis from each laboratory showed relatively low variability; % RSD were typically below 10%, indicating suitable method precision. Carbonyls were the exception, with variability as high as 31% for formaldehyde. This was expected given that these analytes are present at low, or less than LOQ levels, for the samples and tend to be problematic for analytical testing due to ubiquity and relative instability.

Quite a variation, as indicated by LR values in Table 3 and the deviation shown in Figure 3, was observed between labs for the tobacco analytes. This was particularly surprising for tobacco analytes with well-established methodology (2), such as nicotine and NNK with LRs as high as 34% and 247%, respectively. For nicotine, Lab 1 used a continuous flow analysis (CFA) method while labs 2 and 3 used chromatography. But a review of the data does not support a bias between Lab 1 and the other labs due to methodology; the differences do not consistently trend high or low for Lab 1. For the same batch of a product sample, one laboratory reported NNK as 4,388 ng/g while another reported 10,164 ng/g. The low % RSDs reported for a single batch of samples by each laboratory for the tobacco analytes support reasonable method precision. However, the relatively large LR values for most tobacco analytes indicate that the variation in results obtained between labs is more likely due to differences in methodology and inherent product variability. Differences in LPR values also lead one to question the source of differences in the span of means for the different laboratories. Although each individual cigar sample was made up of cigars from the same lot or at least from the same box, product variability appears to be inherent for cigars, especially long filler, hand rolled cigars. There have been considerable efforts towards method standardization, particularly by the CORESTA Tobacco and Tobacco Product Analysis (TTPA) Sub-Group. A review of recent proficiency and collaborative studies did not support that method differences were the likely concern with the spread in labs’ results for this study. There was a recent product characterization study with the newly released cigar reference products that was useful in understanding the tobacco results in this study (32). Participating labs provided 8–13 data sets for a range of tobacco HPHCs and characterizing analytes. Results were used to estimate a true mean for each reference product, and z-scores were calculated for the labs. Generally, laboratories had positive z-scores (< 2), indicating agreement of results. Of note, based on ammonia, TSNA, and pH results, the 1C1 reference product data appeared to be bifurcated, and additional testing confirmed these findings.

Figure 3.

Lab comparison: Tobacco nicotine average of 7 replicates shown for each sample from three different labs. Error bars are 3 × standard deviation.

Analyzing the TTPA study data in the context of this study was enlightening. The laboratory range of the TTPA data (calculated as the percent difference between the highest and lowest reported values) is consistent with the LR results for this study. Nicotine varied by as much as ~30%, and other analytes by 50–80%, excluding 1C1 comparisons. The bifurcated 1C1 results differed as much as 120% between the high and low reporting laboratories. There were sufficient participating labs to conclude that even with this spread in data, the labs’ results were considered consistent based on z-scores. Also, the range in analyte results across the reference samples was similar to this study’s diverse product set, and cigars of a similar design between the two studies (e.g., 1C2 reference product and Sample A) had similar results. This last point supports the value of the reference set to a broad product portfolio.

While we purposely selected products in this study from single lots of product, we have learned from Lindegaard’s research even for samples produced from tobacco grown from single lots of seed and grown in the same locale, finished product cigar tobaccos may produce very different results (33). Arsenic, NNN, and nicotine varied by 379%, 115%, and 11%, respectively. Lindegaard’s results are definitive confirmation of the inherent variability of the tobacco itself.

From the results of this study, taken in context with the work of other researchers and considering a long-term historical analysis of cigarette tobacco (34), it is clear that cigar characterization using tobacco constituent analysis will require ongoing analysis of reference products like the ‘RT’ or ‘1C’ series to gain a grasp on the practical relevance of differences observed among data sets for the same product. For example, for a given analyte, 1,000 ng/g vs. 10,000 ng/g may be within the expected range of a given product.

Overall, the difference in LPR values for the laboratories ranged from 11 (nicotine) to 187 (arsenic) indicating differences in discernability among the labs for some of the tobacco analytes, either due to product variability, even from the same lot, method differences among the labs, or a combination of both.

Requirements for preparing cigars for smoking have a foundation in cigarette preparation requirements with differences attributable to the unique features of cigars. Cigars must be stored under specified temperature and humidity to allow for stabilized weight. The conditioning period is set to a minimum of 72 hours with no maximum time specified. Due to size, some cigars’ conditioning may require as long as multiple weeks to reach weight stabilization. Cigars with volatile additives may be unable to weight stabilize, and an analyst must use judgment when to discontinue the equilibration session. In this study, one of the labs reported issues with weight stabilization for the highly flavored sample with a request for instructions. Cigars with a closed mouth-end must be cut prior to smoking. Cutting distance is specified but may need judgment in cases where the mouth-end is not symmetrical or where a wrapper is held on by a glue spot along the cutting line. Additionally, cutting tools for analytical purposes are not standardized. These examples of judgment may contribute to analytical variability but to a lesser extent than product variability and the variability due to the use of non-standardized methods.

Machine smoking of combustible tobacco products under standardized regimes is for comparative purposes and is not intended to represent the range of consumer smoking behaviors. Thus, it is important to employ standardized equipment, reference products, and methodology to allow the comparison of different products under a common set of controlled conditions. For conventional cigarette products, this standardization process began at least as early as the 1960s and is ongoing through organizations such as CORESTA and ISO Technical Committee ISO/TC 126, among others (35). There are multiple standardized smoking regimes (36, 37) as well as several reference and monitor products designed to support conventional machine-made cigarettes.

Regarding cigar smoking, the product category is relatively broad for design which has created a challenge for standardizing smoking parameters. Industry experts designed a regime based on maintaining a constant puff velocity to allow comparison of products with different diameters (< 8 mm to > 25 mm) and to minimize self-extinguishment during machine smoking (38). The machine smoking parameters employed in this study are shown in Table 4.

The methods for determination of TNCO (“tar”, nicotine, carbon monoxide) are the only smoking methods for cigars to date to be regularly evaluated in lab-to-lab comparisons. The methodology has been used since 2005, with at least 16 completed lab comparison studies to establish and evaluate reproducibility and repeatability (r and R) values. Thus, TNCO should be the best-case scenario for smoking analytes for a lab-to-lab (LR) comparison of smoking and smoke constituents.

TNCO results are shown in Table 5 as mean ± standard deviation (% RSD) of 20 replicates for puff count, CO, nicotine, “tar”, TPM, and water where a replicate is a pad/port count without regard to the number of cigars smoked per pad/port. LR values indicate the range between the different laboratories’ results. As expected based on physical parameters (e.g., tobacco weight), the range (LPR) in puff count among the products was incredibly wide at 20–200 puffs/cigar. The spread in values among the other measures is similar to the puff count trend. The spread in values among the other measures follows the trend of increasing analyte delivery with increasing puff count. The difference in LPR values for the laboratories ranged from 40 to 150 indicating less consistency in results among the laboratories for the products compared to physical parameter measures.

The variability in replicates for the measures is much higher than observed for tobacco measures. This would be expected due to increased complexity for smoking: the addition of a wide range of sample preparation variables related to smoking and the, typically, more complex clean-up and chromatographic analysis of the resulting extracts. Sample A would be considered the easiest to smoke based on size, shape, and relative yields, and it tended to have the lowest % RSD among the samples. The % RSD for water was strikingly high, particularly for Lab 3. Water is a problematic analyte due to its ubiquitous nature, and is formed during combustion processes. The water yields for cigars are relatively high; in our work, we have seen water breakthrough to the smoking machine tubing and CO collection bags.

Sample D, which had a non-cylindrical and tipped mouth-end, had the highest LR in results at 33%–224% for TNCO and water. This was anticipated due to the non-cylindrical mouth-end being more challenging to seal in the cigar holder and is likely due to differences in handling practices and management of the cigar-cigar holder interface. In practice, laboratories use the commercially available holders described in CRM 64 with adaptations for the mouthpiece or custom-made holders of unknown design for samples with non-cylindrical mouth-ends. Lack of standardization and potentially lack of experience with such products may lead to high LR values.

Given the relatively high replicate variability and LR results, a particular question for TNCO results from the study is: Given the range of design features, are the samples distinguishable from each other based on TNCO results? This was not the case in this study. Figure 4 shows an ANOVA comparison for the samples’ “tar” using the data from one laboratory. The vertical span of the diamonds represents the 95% confidence interval for the mean, and the line near the center represents the mean “tar” value for each sample. Only Samples E and F are statistically different from the other samples when compared pairwise using Tukey-Kramer analysis. It is notable that Samples E and F show a very large spread in replicate values. Reproducibility and repeatability results from the 16^th annual TNCO collaborative study conducted by CORESTA support the lack of statistical significance. %R values determined from the results of 7 laboratories (n = 5) ranged from ~30% to 70% for products similar to Sample C in this study (39).

Figure 4.

Product comparison: “Tar” yields. Data shown for a single laboratory (Lab 1), n=20.

Research has shown that laboratories with significant experience testing relatively less-challenging cigars have demonstrated improvement in data agreement when using standardized methodology for cigar smoking. For example, Figure 5 shows a trending improvement in %R values across 11 years of collaborative TNCO testing (40) but also highlights relatively low participation compared to collaborative testing of tobacco and other products.

Figure 5.

TNCO Collaborative Studies: %R values for “tar” from 2007 to 2019 (n=5; laboratory count, typically <12).

Interestingly, prior to the release of cigar reference products, the TNCO collaborative studies were conducted using cigarette monitor test pieces as a study control of sorts. As displayed in Figure 6, with means of a similar magnitude, coefficient of variability (CV) of r and R values between a non-intense cigarette regime and the cigar-smoking regime are similar (40, 41). This is encouraging regarding the cigar-smoking method’s capabilities but points to the relative difficulty for testing and inherent variability of some samples as the prominent cause of higher variability among the TNCO data in this study.

Figure 6.

TNCO Collaborative studies: Cigar (40) vs. cigarette (41) regime for a cigarette monitor test piece where global mean is based on data from all laboratories in the study.

A drive toward simplifying and harmonizing techniques and acknowledging that some cigar designs are not particularly amenable to machine smoking may be warranted for product characterization within and between laboratories. As with analysis and interpretation of tobacco, considering a given dataset in context is critical to decision making.

Methods in use for smoke constituents are typically based on established cigarette methods with necessary modifications for cigars typically due to increased yield and smoke times. These modifications have been made in isolation rather than through collaborative studies. Thus, it is likely that there are many unknown but relevant differences between different labs’ methods.

Results for ammonia, and selected carbonyls, PAAs, PAHs, TSNAs, and VOC are shown in Table 6 as mean ± standard deviation (% RSD) of seven replicates for selected smoke constituents along with LR values for the analytes. For constituent evaluation, data are consistently high for % RSD and LR values. The spread in values between the products (LPR) is approximately 10- to 30-fold.

Ammonia and carbonyl results are particularly interesting in the dataset. In addition to high replicate variability for ammonia, there was no consistency between the three labs. For example, Sample D results ranged from < LOQ to over 100 µg/cigar between the labs’ datasets. Additionally, the ammonia yields for samples E and F are significantly higher than those for samples A–D, suggesting possible ammonia formation during smoking due to the difficulty and length of time required to machine-smoke larger cigars. Research by Prepelitskaya et al. confirms the likelihood of artificial generation of ammonia under certain experimental conditions and the breakdown of ammonia under other conditions. They also confirmed that typical cigarette methodology is not applicable to all cigars (42, 43). Given the results in this study and the background research of Prepelitskaya, the reported ammonia results here would be considered as suspect.

It was anticipated that carbonyls, formaldehyde in particular, would be the most problematic for methods development and testing of cigars. Carbonyl methods involve time-sensitive derivatization that requires optimization of pH and the total amount of derivatizing agent. A direct adaptation of a cigarette method is unlikely to be successful for cigars due to smoking times that may go into hours, compared to approximately 10 minutes for cigarettes. Besides variability and range issues across the labs’ results, the trend for each lab across the samples is notably different for formaldehyde, potentially elucidating methodological issues akin to those suspected for ammonia. As displayed in Figure 7, Lab 2 results trend up across the entire sample range (A–F). At the same time, Lab 1 and Lab 3 appear to have a similar trend in data for A–D but differentiate from Lab 2 for E and F. Crotonaldehyde results for Sample B appear anomalous for Lab 2 in particular, but with no assignable cause for the striking difference in the data reported. There is no consistency among the acetaldehyde or acrolein data reported by the three labs. The difference in LPR values for the laboratories ranged from 47 to 143 indicating less consistency in results among the laboratories for the products compared to physical parameter measures. This information and the relatively high % RSD and LR values for the smoke constituents, reinforce continued methods refinement and use of reference products to aid data analysis and comparison between studies.

Figure 7.

Lab comparison: Smoke formaldehyde average of 7 replicates shown for each sample from three different labs. Error bars are 3 × standard deviation.

The study’s objective was an initial assessment of product range and variability and laboratory range (i.e., variability) for single batch analysis of selected cigars of typical design. Using only one batch of product does not allow for evaluating product batch-to-batch variability. Research akin to the CORESTA Cigarette Variability Task Force (CVAR) study for cigarette batch-to-batch variability (44) to understand cigar product variability, though likely much more challenging than similar undertakings for other product categories, is in progress (45). The number of laboratories in the study does not allow for the determination of z-scores, outliers, or repeatability and reproducibility values. It may or may not be appropriate to compare tobacco nicotine (total alkaloids) and smoke nicotine (nicotine) due to differences in analytes measured.