The Reliability and Stability of Puff Topography Variables in Non-Daily Smokers Assessed in the Laboratory

Abstract

Introduction:

Puff topography variables, often measured using the Clinical Research Support System device, have traditionally been studied in regular, daily smokers and have been shown to be highly stable. However, more recent research has focused on non-daily smokers as a population of interest. As such, the aim of this article was to examine puff topography stability (cross-cigarette agreement over time) and reliability (within-cigarette consistency) in non-daily smokers across six laboratory sessions.

Methods:

One hundred seven non-daily smokers attended six laboratory sessions over the course of 3 months. At each session, they smoked one cigarette through the Clinical Research Support System pocket, in addition to completing questionnaires about their smoking history and dependence.

Results:

Puff topography measurements were highly reliable (α values ranged from 0.87–0.95) and puff behavior was highly stable across sessions ( r values ranged from 0.38–0.84). Adding sessions substantially improved reliability estimates. Aspects of puffing behavior observed in session, including puff volume, puff duration, time of puff peak, and total cigarette volume were related to level of smoke exposure, measured by expired carbon monoxide. Instability in puffing behavior was not predicted by recent or long-term smoking patterns.

Conclusions:

Puff topography appears to be a stable and routinized aspect of smoking in non-daily smokers. The feasibility of assessing puff topography in this population is supported by the high reliabilities observed, though it should be noted that reliability greatly improved by having more than one session.

Introduction

The majority of what is known about smoking behavior has come from research on heavy, regular smokers, the majority of whom smoke daily. Smokers are at substantial risk for a myriad of preventable smoking-related diseases, including heart disease, cancer, and emphysema, ¹ and research efforts have focused on understanding both clinical and pre-clinical processes in these smokers. National surveys, however, have revealed that the number of non-daily and non-dependent smokers appears to be rising; 22%–33% of adult smokers in the United States do not smoke daily. ² The main criterion for categorization as a non-daily smoker includes smoking on at least one, but not all, of the past days of an assessment period and can encompass a wide range of smoking levels. Non-daily smokers are also at increased risk for smoking-related diseases and at about one and a half times the risk of early mortality compared to nonsmokers. ³ Given the high and increasing prevalence of non-daily smokers, as well as the health risks associated with this level of cigarette use, research on these smokers is increasingly vital.

Daily smokers are generally thought to regulate their smoking behavior throughout the day with the predominant goal of avoiding nicotine withdrawal symptoms. ^{4 , 5} Less is known about smoking patterns in non-daily smokers; recent research has shown that they are likely to smoke socially, when experiencing negative affect, and in specific situations where they usually smoke (eg, after eating). ^5–7 These studies demonstrate that there is some predictability in non-daily smokers’ patterns of smoking—their smoking choices do not appear to be random. Non-daily smokers experience several phenomenon that have predominately been studied in daily smokers, including generalized (or background) craving, ⁸ cue-specific craving, ⁹ and motivation to cut-down on smoking, ¹⁰ though the magnitude of these effects may vary as a function of smoking level.

The aforementioned studies have examined several features of smoking behavior, but have not addressed the reliability of how a non-daily smoker might smoke his or her cigarette. We know relatively little about the smoking patterns of non-daily smokers within cigarettes, that is, puff topography characteristics including number of puffs, inspired puff volume, and the inter-puff interval. Puff topography variables have displayed extremely high stability in regular daily smokers across cigarettes in a single session ^{11 , 12} as well as across multiple sessions, ¹³ though there is substantial variability in puff topography between smokers. ¹⁴ This smoking topography pattern has generally been shown to be robust under variable conditions, ^{15 , 16} and puff topography characteristics have demonstrated high reliability. ¹³ Evaluating the reliability and stability of smoking characteristics in a non-daily sample will help researchers understand more about this understudied group and could also provide information about smoke toxicity in this sample.

To date, the reliability and stability of smoking topography has not been evaluated in non-daily smokers. The first aim of the current study was to evaluate the reliability and stability of several smoking topography variables in a sample representing a wide range of non-daily smoking. To clarify, we are using reliability to measure consistency of puffing characteristics within-cigarette across multiple sessions, and stability to measure cross-cigarette agreement for each topography variable over time. A second aim was to evaluate the reliability of smoking topography across multiple sessions with the goal of predicting the reliability of a single session of smoking topography. The majority of studies that have analyzed smoking topography typically only used data collected from one cigarette ¹⁷ ; single measurements, however, suffer from diminished reliability compared to multiple measurements. ¹⁸ Some theories of addiction purport that drug-use behavior should become automatized over time and repeated drug-use exposures; ¹⁹ therefore, greater stability in smoking topography may be seen in those who smoke more frequently. To test this theory, in the third aim we examined whether smoking topography stability could be predicted by nicotine dependence or recent smoking history (ie, cigarettes smoked per day in the last month and days smoked in the past month). Finally, given the health risks associated with non-daily smoking, we were interested in analyzing the relationship between puff topography and carbon monoxide (CO) levels, a biomarker of smoke exposure.

Method

Participants were 107 non-daily adult smokers recruited by newspaper/radio advertisements and flyers from the Western New York area. Individuals were pre-screened via telephone to determine eligibility. These participants were a subset of participants recruited for a larger study on hair toxicology in both daily and non-daily smokers. Participants retained from the larger study must have attended all six sessions and could not have smoked 28 of 28 days on the Timeline Follow-Back given at Session 1. Furthermore, only participants with Clinical Research Support System (CReSS) data from all six sessions were used in the current study. Eligible participants were between 18–45 years of age and smoked between 1–29 of the last 30 days with no more than 15 cigarettes on any given smoking day and had to provide an expired pre-cigarette CO level no greater than 15 parts per million at their first study visit. These criteria were established with the goal of producing a widely heterogeneous sample so that we could explore any patterns attributable to smoking level in a continuous rather than categorical manner, which we believed would provide a more fine-grained analysis. Participants also had to have smoked at least 25 cigarettes in their lifetime and must not have been interested in quitting or cutting down on smoking during the 3-month study period, and could not use any other tobacco products. Due to requirements of the larger study, participants were excluded if they had predominately gray or white hair, if they had hair shorter than 1.25 inches, and if they were unwilling to forego hair-coloring treatment over the next 3 months.

Eligible participants attended six study sessions over the course of 3 months; sessions occurred at the same time every week for the first five sessions (Sessions 1–5) and the final session (Session 6) occurred 8 weeks after Session 5. Sessions ranged in length from 60–120 minutes. Participants were paid up to $300 for completion of all study sessions. A variety of tasks were completed during each session; only those procedures relevant to the current study will be described.

At Session 1, participants were consented and then provided expired breath CO through a Vitalograph BreathCo Carbon Monoxide Monitor. Self-report measures were completed including demographics, smoking history, the Fagerström Test for Nicotine Dependence, ²⁰ and the Wisconsin Inventory of Smoking Dependence Motives (WISDM-68). ²¹ Participants then smoked one of their own cigarettes using the handheld CReSS device (borgdwalt.ce), which records several measures of smoking puff topography (see below). Participants who did not provide their own cigarettes were given a choice of cigarettes from the laboratory. Participants were then interviewed using 1-month Timeline Follow-Back ²² to assess their past 28-day smoking behavior. At Sessions 2–6, participants provided a CO sample and smoked one cigarette through CReSS. At Session 5, participants completed an additional Timeline Follow-Back to account for smoking behavior during the first 28 days of the study.

Data Reduction and Analysis

For each cigarette smoked, puff-level data were averaged for each of the following CReSS-produced variables: average puff volume (measured in milliliters), average flow rate (the average flow over the course of the puff; measured in milliliters per second), peak flow rate (the peak flow rate that occurred over the course of the puff; measured in milliliters per second), puff duration (the duration of the puff; measured in milliseconds), interpuff interval (the time between the end of the previous puff and the start of the next puff; measured in milliseconds), time of peak (the time between which the puff began and the point at which the peak flow was detected), cigarette duration (the time elapsed from when a person began the first puff of the cigarette and ended the last puff of the cigarette; measured in milliseconds), total puff volume (the sum of each puff’s volume; measured in milliliters) and number of puffs taken.

A participant’s data was removed from analyses when the CReSS device appeared to malfunction (eg, if CReSS produced only a single puff when participants were observed to puff more). In relatively few instances, a single puff was removed from analysis but the rest of that cigarette was retained if CReSS appeared to malfunction for only one puff. This was limited exclusively to the first or the last puff of the cigarette. In all, 18 participants were removed from analyses because multiple variables from their cigarettes were questionable (the participant count, above, reflects the number of participants analyzed after these cases were removed) and six individual puffs were removed (these participants were retained in analyses). All CReSS data were reviewed by two independent coders who achieved very high agreement regarding case removal (>99%).

Aim 1

Cross-cigarette stability was assessed for each topography variable by calculating correlation coefficients for adjacent sessions. That is, Session 1 was correlated with Session 2, Session 2 with Session 3, and so on, resulting in five correlation coefficients per variable. Cronbach’s α was also calculated across the six sessions to estimate the reliability of each smoking topography variable derived from aggregated sessions. ²³

Aim 2

Given that researchers typically obtain measurements of puff topography for only one or two smoked cigarettes, we wanted to evaluate the effect of number of smoked cigarettes on reliability of data aggregated across multiple cigarettes; thus, α was also calculated for the first two, three, four, and five sessions of smoking. The Spearman–Brown Prophecy Formula, ^{24 , 25} with α calculated from six sessions, was used to estimate the reliability of topography variables from a single session.

Aim 3

For each puff topography variable, an instability index was calculated for every participant. This index was calculated by determining each participant’s average score across the six sessions for every variable, then subtracting that average from the data of each of the six individual sessions. This resulted in six “deviations” from a participant’s average behavior. These values were then squared and averaged, resulting in an index wherein higher values represented greater instability and lower values represented relative stability. This variable was used in regression analyses to test the hypothesis that smokers exhibit more stability in their smoking behavior as they have more experience with smoking. These instability indices were entered as dependent variables in two-level sequential multiple regression equations with continuous predictors and outcomes. In Step 1, dependence scores (measured by the WISDM-68) and years smoking were entered, both variables that signified longer-term patterns of smoking behavior. In Step 2, variables more reflective of recent smoking behavior over the course of the study were entered: average cigarettes per day over the first 28 days of the study and average CO over the course of the study.

Aim 4

To assess the correspondence of puff characteristics with smoke exposure, one representative puff topography variable was selected from three categories. Unpublished research from this laboratory suggests that when puff topography variables are factor analyzed, three factors are extracted: inhalation (which included indices of volume, time to peak, and puff duration), flow (peak flow and average flow), and cigarette pacing (puff number and interpuff interval). Total puff volume and average cigarette duration do not appear to uniquely belong to any single factor. In the current research, because entering all nine topography variables into one regression model would reduce our degrees of freedom, we decided to include one variable from each factor that had the strongest factor loading. The variables selected had standardized factor loadings as follows: puff duration (0.96), puff number (0.93), and average flow (0.93). These variables, and their interactions, were used in regression analyses to predict aggregated pre-cigarette CO over the six sessions. For these analyses, all variables were standardized to a mean of 0 and standard deviation of 1. The first model only considered main effects of each variable, and the second considered main effects and interactions of all variables.

$$\begin{array}{l}Y\left(CO\right)=a\left(\mathrm{int}ercept\right)+{\beta }^1\left(puffnumber\right)\\ +{\beta }^2\left(flowrate\right)+{\beta }^3\left(puffduration\right)+e\end{array}$$ (1)

$$\begin{array}{l}Y\left(CO\right)=a\left(\mathrm{int}ercept\right)+{\beta }^1\left(puffnumber\right)+{\beta }^2\left(flowrate\right)\\ +{\beta }^3\left(puffduration\right)\\ +{\beta }^{12}\left(puffnumber\times averageflowrate\right) \\ +{\beta }^{13}\left(puffnumber\times puffduration\right)\\ +{\beta }^{23}\left(averageflowrate\times puffduration\right)\\ +{\beta }^{123}\left(puffnumber\times averageflowrate\times puffduration\right)\\ +e\end{array}$$ (2)

Results

Data from 107 participants (57% female, 57% white) was analyzed. Participants averaged 25.0 years of age and had been smoking for 9.0 years. See Table 1 and Figure 1 for a summary of participant characteristics.

Table 1.

Participant Characteristics and Smoking Behavior Profiles

	Mean ( SD )	Range
Age a	25.0 (5.9)	18–45
FTND a	1.1 (1.5)	0–5
WISDM-68 a	32.1 (13.3)	13–73
Average CO level c	5.2 (5.8)	0.4–30.4
Days smoked in last 28 during study period b	17.7 (7.8)	4–28
Average cigarettes per smoking occasion during study period b	2.7 (2.2)	1–14.5
Years smoking a	9.1 (7.1)	0–28.0
Days since most recent cigarette c	2.0 (2.9)	0–22.0

CO = carbon monoxide; FTND = Fagerström Test for Nicotine Dependence; WISDM-68 = Wisconsin Inventory of Smoking Dependence Motives.

^a Measured at Session 1.

^b Measured at Session 5.

^c Measured at all sessions.

Figure 1.

Participants categorized by number of cigarettes smoked in the last 28 days.

Aim 1 and 2—Within-Cigarette Reliability and Cross-Session Stability of Smoking Topography Variables

In general, across the nine smoking topography variables, very high reliability and stability were achieved across the six sessions ( Table 2 ). Cronbach’s αs ranged from 0.87–0.95, indicating very high within-cigarette reliability across the six sessions. Overall, puff duration was the most reliable variable (α = 0.95) and total cigarette volume was the least reliable (α = 0.87). Aggregating data across sessions, as expected, led to increased reliability ( Figure 2 )—reliability decreased by up to 25% when only two sessions of data were considered. The Spearman–Brown Prophecy formula suggested that having only a single session of puff topography data further reduced reliability to estimated αs of 0.53–0.76, which was up to a 40% decrease in reliability compared to six sessions of aggregated data. Inspection of the reliability estimates by number of sessions contributing to the estimate ( Figure 2 ) revealed that reliability for puff topography leveled-off at four sessions of data for most of the variables. For all nine variables, all adjacent session correlations were significant at P < . 0001, with Pearson’s r s ranging from moderate to high (0.38–0.84) with an average cross-session correlation of 0.65 ( Table 2 ). Puff duration displayed the strongest session-to-session stability (average r of 0.78), and total volume displayed the lowest level of session-to-session stability (average r of 0.54). Averaged across all nine variables, the session-to-session correlations remained highly consisent across ordered session pairs (range 0.63–0.67).

Table 2.

Means, Standard Deviations, Cronbach’s α, and Session-by-Session Pearson’s Correlations for all Puff Topography Variables

Variable	Mean ( SD )	Cronbach’s α of 6 sessions	Predicted α of a single session	r at Sessions 1 and 2	r at Sessions 2 and 3	r at Sessions 3 and 4	r at Sessions 4 and 5	r at Sessions 5 and 6
Number of puffs	13.5 (4.5)	0.89	0.59	0.57*	0.49*	0.38*	0.65*	0.72*
Average puff volume (mL)	31.4 (13.0)	0.93	0.69	0.75*	0.72*	0.73*	0.72*	0.63*
Total puff volume (mL)	413.6 (182.2)	0.87	0.53	0.50*	0.60*	0.46*	0.50*	0.63*
Puff duration (s)	1.2 (0.5)	0.95	0.76	0.72*	0.69*	0.84*	0.83*	0.80*
Cigarette duration (m)	5.6 (1.6)	0.89	0.56	0.63*	0.69*	0.55*	0.49*	0.52*
Puff flow rate (mL/s)	26.8 (7.9)	0.91	0.64	0.69*	0.79*	0.70*	0.70*	0.55*
Peak flow rate (mL/s)	34.9 (11.8)	0.92	0.64	0.72*	0.79*	0.69*	0.71*	0.57*
Interpuff interval (s)	28.8 (45.4)	0.91	0.63	0.59*	0.69*	0.73*	0.65*	0.59*
Time of peak (s)	0.5 (0.2)	0.88	0.54	0.48*	0.42*	0.78*	0.82*	0.73*

*Signifies P < .001.

Figure 2.

Cronbach’s α as a function of the number of sessions contributing to the average score for each puff topography variable. The Spearman–Brown Prophecy Formula was used to estimate Cronbach’s α for 1 session.

Aim 3—Predicting Instability

Models predicting puff topography instability with dependence level, years smoking, average CPD, and CO did not approach significance (all P s > .30).

Aim 4—Predicting Smoke Exposure

Aggregated pre-cigarette CO measurement across the six sessions was significantly and positively predicted by the main effects regression model, F (3, 102) = 3.66, MSE = 0.93, P = .015, η ²_p = 0.1, r²= 0.10. These effects were driven by the significant relationships between average flow and pre-cigarette CO ( t = 2.23, P = .028, r²= 0.03) and average duration and pre-cigarette CO ( t = 2.77, P = .007, r²= 0.07). Standardized beta coefficients suggest that a 0.2 unit increase in flow corresponded to a 1-unit increase in CO, and a 0.3 unit increase in puff duration corresponded to a 1-unit increase in CO. Puff number did not significantly predict CO ( P = .50). The second regression model was not significant overall ( P = .13), and no interaction between topography variables significantly predicted pre-cigarette CO.

Discussion

In this study, non-daily smokers demonstrated high cross-cigarette stability and reliability of measurement of puff topography across six smoked cigarettes. This is the first study to measure the stability and reliability of puff topography in non-daily smokers across multiple sessions. Puff topography instability in this sample did not appear to vary as a function of smoking level (both measured by recent smoking and longer-term smoking patterns), suggesting that the distinct puff topography pattern readily observed in regular, daily smokers is also evident in non-daily smokers. Further, smoke exposure appears to be related to several characteristics of how non-daily smokers puff their cigarettes.

The overall reliability of these puff topography variables was very high, suggesting that topography variables obtained via CReSS can be measured with precision in non-daily smokers when the variables are aggregated over multiple cigarettes. The high reliabilities across variables support the use of puff topography assessment in this population. However, reliabilities declined sharply as the number of sessions decreased ( Figure 2 ); future researchers should consider multiple assessments of puff topography in non-daily smokers to optimize reliable measurement. Per our results, for most variables it appeared that four smoked cigarettes were sufficient to generate optimal reliabilities (the addition of cigarettes 5 and 6 produced only small increments in reliability).

The overall pattern of puff topography between cigarettes in these non-daily smokers was highly stable. Session-to-session correlations demonstrated that even these less-practiced smokers tended to smoke their cigarettes with a high degree of stability. This finding adds to the growing literature on non-daily smokers—not only do non-daily smoking individual tend to smoke under somewhat stable circumstances, ^{4 , 6 , 7} their puffing behavior is also stable. Our results suggest that puff behavior is a distinct feature of the behavioral repertoire of smokers that may routinize early in a person’s smoking history.

At the lower end of our sample were rather inexperienced smokers (~1 cigarette per month); however, smoking characteristics (recent or long-term) did not generally predict puff topography instability. Thus, in our sample, even the most inexperienced smokers had already established a puff topography pattern. This finding may lend support for Tiffany’s theory of drug-use automaticity ¹⁹ in that there appear to be aspects of smoking that become automatized quite quickly. The non-daily smokers in our study had day-to-day variability in their smoking frequency; however, their style of puffing when they did smoke was highly invariable, suggesting that patterns of smoking within cigarettes (ie, puff topography) may be more rapidly automatized than smoking across cigarettes (ie, patterns of smoking frequency over days). Our instability results may be affected in part by the screening criteria that participants had to smoke at least 25 cigarettes in their lifetime; individuals smoking the first few cigarettes of their lifetime are unlikely to have yet developed stable puffing behavior. In addition, our regression models were adequately powered to find medium or large effect sizes, but we were unable to detect a small effect size due to our sample size; thus, our sample size may have inhibited these results as well as the regression models (Aim 4) predicting pre-cigarette CO.

In our study, puff topography was quite stable over the six sessions. The six sessions, however, were highly similar and thus participants likely adjusted quickly to the lab environment. Though this allowed for high experimental control, our study did not specifically measure how puff behavior might change under variable conditions. In daily smokers, puff topography changes have been observed under different conditions, including the presence of mortality cues, ²⁶ the presence of alcohol cues, ¹⁷ and while under the influence of alcohol. ²⁷ Thus, it may be informative to assess when and under what conditions non-daily smokers may alter their puffing behavior and whether the conditions that affect puff topography differ as a function of smoking experience.

Characteristics of puff topography, specifically those representing flow and inhalation, appeared to predict pre-cigarette CO in this sample of non-daily smokers, though the overall variance in CO accounted for was low (7%). This finding suggests that the manner in which a person smokes affects the amount of smoke they are exposed to from that cigarette—and that this puff behavior may be associated with negative health outcomes. Our analyses revealed that though number of individual puffs does not appear to correspond to smoke exposure, it is the individual characteristic of the puff that plays a role (specifically, inhalation, and flow variables). As researchers become increasingly interested in non-daily smokers and begin to understand more about the patterning of their smoking in the natural environment, ^{4 , 6 , 8} the simultaneous assessment of puff topography may provide additional insights into smoking choices. Our study took place exclusively in the laboratory—in order to generalize these findings, future research could focus on examining the stability of puff behavior in the natural environments of non-daily smokers. Because CO was measured only once at each session, and was measured pre-cigarette, our results may suffer from reduced variability and precision in CO measurement. Future researchers may wish to look at the relationship between topography and post-smoking CO boost as a more accurate measure of the impact of topography on immediate smoke exposure.

In summary, this was the first study to look at the reliability and stability of puff topography measured via CReSS in non-daily smokers over six sessions. Our findings support the use of puff topography assessment in this population. These results also suggest that even though non-daily smokers may not be as routinized as daily smokers in their day-to-day smoking frequency, how they puff is a unique and stable characteristic of their smoking behavior.

Funding

This work was supported by the National Cancer Institute (grant number R01 CA120412).

Declaration of Interests

None declared .