Recruitment activities were conducted by Battelle Memorial Institute. Recruitment efforts began on November 24, 2003 and continued through April 28, 2004. Inclusion criteria were: self-identification as African American or Caucasian; 21 years of age or older; established smokers, defined as smoking daily, at least 6 cigarettes per day, and smoking for at least 3 years; flexibility in smoking unfamiliar brands and willingness to smoke both menthol and nonmenthol cigarettes; and ability to attend 3 visits, each lasting approximately 2 hours. Pregnant participants, participants arriving intoxicated to any visit, participants with self-reported smoking-related diseases, or users of the test cigarettes were excluded. Recruitment efforts initially consisted of advertising in local newspapers and posting flyers in the Baltimore, MD area. An incentive word-of-mouth reimbursement program and recruitment at bingo parlors, bus stops, malls, bowling alleys, and designated smoking areas were later implemented to increase recruitment of African American nonmenthol smokers. In the word-of-mouth reimbursement program, participants who referred others were eligible for a $10 incentive for each eligible “hard-to-reach” participant that came to their first appointment. This resulted in an increased screening volume. As reported in the literature,⁴ African Americans disproportionately smoke menthol cigarettes and African American smokers of nonmenthol cigarettes were hard to reach participants in the recruitment area.

Participants were asked to come to the Smoking Research Laboratory of Battelle Centers for Public Health Research and Evaluation in Baltimore, MD for the 3 required laboratory visits and prior to data collection, each prospective study participant met in person with the Battelle project coordinator. At this meeting, the coordinator explained the study procedures and answered the participant's questions. The study participant reviewed the informed consent form with the project coordinator. If the participant agreed to participate in the study, (s)he was asked to sign the consent form. The study protocol was approved by institutional review boards at CDC and Battelle.

Participants were randomly assigned to alternate between 2 weeks of exclusively smoking the menthol test cigarette or the nonmenthol test cigarette. The study design consisted of 3 laboratory visits. There was no orientation visit. Protocols for laboratory visits 1-3 are depicted in Figure 1. When participants arrived for their first visit, they read the consent form and filled out a detailed smoking history questionnaire. The questionnaire covered current smoking behavior, brand preference, smoking history such as age at first cigarette, interest in quitting, and self-reported measures related to addiction including “how soon after you wake up do you smoke your first cigarette?” Self-reported cigarettes smoked per day and time to first cigarette from the smoking questionnaire were used to calculate “heaviness of smoking” scores.²³ Baseline urine, saliva, and carbon monoxide (CO) were collected. Participants familiarized themselves with the CreSS® smoking topography device (Borgwaldt-KC, Richmond, VA) by smoking one of their own cigarettes using the device. Two minutes after smoking their own cigarette, another breath CO level was measured. After 30 minutes breath CO was measured again to ascertain that participant was not too satiated from the previous cigarettes and the participant smoked the randomly assigned menthol or nonmenthol test cigarette using the CreSS® device. Topography recording began following the “lighting puff.” Carbon monoxide levels were measured again 2 minutes after smoking the test cigarette. This concluded the first visit and participants were provided with the test cigarettes, collection bags, and instructions for collecting each day's cigarette butts in individual bags (Day 1 - Monday, Day 2 – Tuesday, etc.). Participants were instructed to only smoke the test cigarettes and to smoke as little or as much as they wanted each day. Participants were provided with what was estimated to be enough packs of the test cigarette to last until their next appointment based on their reported number of cigarettes smoked each day. Exposure to other sources of menthol was limited by providing nonmentholated toothpaste and requesting that participants refrain from mentholated products (gum, mints, and candy) while in the study.

After 2 weeks participants arrived for their second visit. Their cigarette butts were collected and counted. If more than 10% of the cigarette butts were non-test cigarette butts, or if there was an excessive number of missing butts, the participant was considered noncompliant and was not allowed to continue in the study. Participants were asked to give a urine sample and 2 saliva samples. The smoking procedure for visit 1 was repeated. Two minutes prior to smoking, participants were asked to give a breath CO sample. The first cigarette smoked using the CreSS® device was the test cigarettes they had smoked during the previous 2 weeks. Two minutes following smoking, breath CO was measured. Participants waited for 30 minutes before smoking the other test cigarette. Instructions for cigarette butt collection were again explained to participants, and participants were sent home with the second test cigarette. Visit 3 was 2 weeks after visit 2 and followed the same protocol as visit 2.

At each visit participants completed a 6 point Likert style survey that rated their impression of the test cigarette in terms of satisfaction, enjoyment, throat irritation, aftertaste, smoke smell, and package smell. All cigarette filter butts and biological specimens were shipped to the Centers for Disease Control and Prevention in Atlanta, Georgia where they were stored frozen at -70°C until analysis. The majority of the analytical work (described below) was conducted between 2009 and 2010.

Ideally test cigarettes would be commercial cigarettes that are chemically and physically identical, differing only in menthol content. In the absence of matched menthol and nonmenthol cigarettes, a nonmenthol cigarette (Kent 100 soft pack, RJ Reynolds Tobacco Company, Winston-Salem, NC) and a menthol cigarette (Benson & Hedges Light 100 soft pack, Philip Morris USA, Richmond, VA) were selected on the basis of calculated Pearson product-moment (parametric) and Spearman rank (non-parametric) correlation coefficients for current and publically available data on mainstream smoke constituent levels from the 1999 Massachusetts Benchmark Study Final Report.²³ All brand combinations were highly correlated, with Pearson correlations >0.90 and Spearman correlations >0.87. The sum of the relative percent differences (RPD) between the constituents for all possible menthol/nonmenthol brand combinations was also examined. The minimum average RPD between a given smoke constituent measured in a menthol versus a nonmenthol brand was found to be 7.9% from the 1999 Massachusetts Benchmark Study and 8.2% from the Swauger study.^23,24 Additionally, participants who reported smoking either of these cigarette brands were excluded to reduce any effect of familiarity. Mainstream smoke characterization of the 2 test cigarettes is published elsewhere.²⁶ Test cigarettes were matched for length, circumference, and tobacco weight. Filter ventilation for the mentholated Benson & Hedges cigarette and nonmentholated Kent cigarette were 28±1.5% and 19±1.2, respectively.²⁶

Case cross-over study designs entail repeated measurements on participants who each undergo all exposures of interest, but at different time periods so that each participant “crosses over” from one type of exposure to another. Since each participant is measured while undergoing the reference exposure, they provide their own baseline measurements for comparison to when they undergo other exposures. The case cross-over design is preferred for studies attributing biomarker response to exposures where the biomarker's presence is transient or reversible so that a participant's biomarker levels can return to baseline before the participant crosses over to the next exposure regimen. The order of exposure is generally designated as the study “sequence” and the time of exposure as the study “period.” Potential bias from carry-over effects on biomarker measurements from previous periods were evaluated in statistical analysis¹⁹ and found to be non-significant. The 2 exposures examined in this study were to smoking nonmenthol cigarettes (designated “S”) or menthol cigarettes (“M”). Participants were randomized to 4 uniform-within-period sequences each spanning 3 periods: MMS, MSM, SMS, and SSM. The unbalanced numbers of menthol preferring and nonmenthol preferring participants were accounted for indirectly by weighting calculations proportionally to the number of each type of participant.

For each biomarker, a mixed effects model was formed with fixed effects for cigarette type, sequence, and period, as well as a random effect for each participant. In addition, urinary metabolite concentrations were transformed with the natural log and the natural log of urinary creatinine was included as a predictor to control for varying hydration. Potential first- and second-order carryover effects were evaluated in the mixed effects models, found to not be statistically significant, and were therefore excluded from the reported models. Mixed effects models were also configured to compare the effect on model fit from assumptions of homogeneous vs. heterogeneous variances among periods and sequences, as well as assumptions about whether within-participant covariances were unstructured or of the compound symmetry type. Based on Bayesian information criterion, the optimal configuration for the mixed effects models comprised homogeneous variances among periods and sequences, with within-participant covariances structured for compound symmetry.²² Reported results, including least-square means and standard errors, are from mixed effects models with this optimal configuration. Statistical analysis was conducted using the PROC MIXED subroutine of the SAS/STAT software application version 9.3 (SAS Institute, Cary, NC) with estimation by restricted maximum likelihood. Sex was included as a model predictor because factors associated with sex are known to influence tobacco smoke biomarker concentrations, among them metabolic processing. As very few African American smokers of nonmenthol cigarettes were included in the final data set it was not possible to look at race and exposure to nicotine and other smoke constituents.

Of the 64 participants who completed the study, 42 participants were deemed compliant and had complete data sets from all 3 visits. Scheduling conflicts and difficulty with transportation to laboratory visits were the most common reasons for participant dropout. Participants were labeled as non-compliant if more than 10% of their collected cigarette butts contained were non-test cigarette butts; if they reported exclusively smoking non-test cigarettes at home; if their carbon monoxide levels were non-detectable implying that they did not meet the inclusion criteria of smoking at least 6 cigarettes per day; or, if they did not appear to inhale while smoking through the topography device. NicAlert Strips (Nymox Corporation, NJ) that measure cotinine in urine were also used at laboratory visits to confirm smoking status.

To be included in the final data set, participants had to have at least one topography measure per test cigarette, complete urine and saliva biomarker data, breath CO data, and to be missing no more than 2 days' worth of collected cigarette butts for each 14-day period. The demographics and smoking history characteristics of the 42 participants included in the final data analysis are presented in Table 1. Study enrollment resulted in the following groups of smokers with a menthol preference: 8 female African American menthol smokers, 6 male African American menthol smokers, 8 female white menthol smokers, and 4 male white menthol smokers. For nonmenthol smokers there were 6 white females, 7 white males, 2 African American females and one African American male.

Menthol and nonmenthol smoker participants reported similar numbers of cigarettes smoked per day, age at first cigarette, and age at becoming a daily smoker. The overall average number of cigarettes smoked per day was 22. Roughly half (54%) of current menthol smokers initiated with a menthol cigarette, while 81% of current nonmenthol smokers initiated on a nonmenthol cigarette. Both menthol and nonmenthol smokers reported that they transitioned from trying their first cigarette to daily smoking over approximately 2 years. Half of menthol smokers indicated that they smoked their first cigarette within 5 minutes of waking rather than at later times. A larger percentage of nonmenthol smokers reported smoking their first cigarettes within 30 minutes (44%) than within 5 minutes (38%). Further, a higher percentage of nonmenthol smokers (75%) than menthol smokers (58%) reported that it would be harder for them to give up the first cigarette of the day than a later cigarette. Finally, 65% of menthol smokers reported smoking while ill, while only 36% of nonmenthol smokers reported the same. Using the sum of 2 self-reported measures of heaviness of smoking³² (number of cigarettes per day and time to smoking after waking), a “heaviness of smoking” scale was calculated. Using this stratification scheme, roughly 30% of both nonmenthol and menthol smokers were classified as “highly dependent” and a similar distribution was seen across groups (Table 2). Distribution of menthol or nonmenthol smokers is similar for the 3 study designated categories of low, moderate, or high dependence with the possible exception of twice as many menthol smokers (12%) that could be considered to be lower in dependence than nonmenthol smokers (6%). However, for smokers or 2 or more cigarettes per day, these calculated measures indicate a similar level of tobacco dependence for both groups of smokers.

At the first visit (baseline), no significant differences were seen in biomarkers of exposure between menthol and nonmenthol smokers when smoking their normal brands (data not shown). When participants smoked the test cigarettes, urinary menthol levels ranged from 0.85 to 18.2 mg/L for the menthol brand and 0.34 to 6.4 mg/L for the nonmenthol brand. Average urinary menthol was significantly higher for most participants when smoking the menthol study cigarette than when smoking the nonmenthol test cigarette, indicating a high level of compliance in switching between brands (p = < .0001). However 4 participants had similar or even higher urine menthol levels when smoking the nonmentholated brand. Although provided with nonmentholated toothpaste, participants may have consumed other mentholated products or could have smoked non-test cigarettes without reporting it, but without confirmation of noncompliance, the participant data sets were not excluded. Average biomarker levels and standard errors (SEM) are presented in Table 3.

Both nonmenthol and menthol smokers had significantly higher salivary cotinine when smoking the nonmenthol test cigarette. When cotinine levels were compared to heaviness of smoking scores (Table 2), cotinine had a significant pattern (p = .01) of increased heaviness of smoking with increased cotinine (controlling for cigarette type, sequence, and study period). There was no significant difference in free or total urinary NNAL (p = .37 and .50), for menthol and nonmenthol smokers when smoking either test cigarette. However, there were significant differences by gender with women having significantly higher (p = .05) total NNAL than men regardless of test cigarette smoked (Table 4). There was no significant difference in CO boost for any test period (data not shown).

There were no significant differences in average number of butts collected per day by test cigarette or preference, with roughly 14-16 cigarettes smoked per day of the study (Table 3). Overall, average MLN was 1.03 mg/cig when smoking the menthol test cigarette and 0.87 mg/cig when smoking the nonmenthol cigarette (p = .02) (Table 3). The average nicotine ratio, defined as per cigarette MLN menthol/ per cigarette MLN nonmenthol, was 1.6. Of the study completers, men had significantly higher MLN and total smoke (mean puff number X mean puff volume) than women when smoking either test cigarette (Table 4). Women had higher levels of total NNAL and free NNAL than men when smoking either test cigarette (Table 4).

There were significant differences in puff volume and duration, with participants taking deeper, longer puffs when smoking the menthol cigarette (Table 5). Men inhaled significantly more total smoke than women when smoking either test cigarette (Table 4). Other topography measures included in this study did not reach significance.

When smoking the test cigarette of their preferred flavor, roughly 25% of both groups rated it enjoyable. Menthol smokers were much more likely to rate the nonmenthol test cigarette unfavorably (not enjoyable 84.6%, unpleasant aftertaste 88.5%, unpleased pack (80.8%) or smoke (80.8) smell) than the menthol test cigarette. Nonmenthol smokers similarly rated sensory attributes of the menthol test cigarette unfavorably. When asked whether the test cigarette was satisfying, similar percentages of nonmenthol smokers reported the menthol (68.8%) and nonmenthol (62.5%) test cigarettes as unsatisfying. In contrast, menthol smokers were twice as likely to rate the nonmenthol test cigarette as not satisfying (84.6%) as the menthol cigarette (38.5%). No nonmenthol smokers rated the menthol test cigarette as satisfying or enjoyable while a few of the menthol smokers found the nonmenthol test cigarette satisfying (3.8%) and enjoyable (7.7%). There were statistically significant differences in subjective ratings for the test cigarettes by menthol and nonmenthol cigarette preference, but not by test cigarette switching pattern. Both menthol and nonmenthol smokers reported significantly lower enjoyment and satisfaction scores for both test cigarettes (p = < .0001). Further, participants reported significantly greater throat irritation (p = .03), worse aftertaste (p = .004), worse pack smell (p = .03) and worse burning smell (p = .006) when smoking the brand opposite their menthol preference.