Sample selection

Participants of this study were selected through their enrollment in a customer loyalty program headed by the S Group, a commercial enterprise possessing more than 45% of the market share in Finland¹⁶. The S Group’s large market share minimizes but does not negate the possibility that study participants could have purchased cigarettes and NRT from other sources. Using a database containing customers’ email addresses, the S Group contacted members of the customer loyalty program in the HOK–Elanto retail co-operative in southern Finland. The email inquiry sought members’ consent to release basic demographic information (age, gender, postal code) and purchases made in the year 2016. Release of the data, which excluded personal identifiers, was intended for research purposes and approved by the ethics committee of the University of Helsinki Review Board in the humanities and social and behavioral sciences. Approximately 5% of members of the loyalty program consented to the data release, yielding a sample of 13274 customers. Compared to the general population in southern Finland, consenting members of the loyalty program were predominantly female (about 68%), middle-aged (about 46 years old), and more educated, which can be attributed to higher participation rates¹⁶. This is evidenced by the statistic indicating that 56% versus 33% of consenting loyalty members and the general population, respectively, had a university degree¹⁷. Our final analytic sample, which excluded S Group personnel, consisted of customers who purchased at least one NRT product in the year 2016 (n=728).

Measures

The measures were the time, date, and expenditures of cigarettes and NRT in the year 2016. Categories of NRT consumers were based on the purchase of one or more products in consecutive 4-week intervals. The categories, which were based on clinical guidelines and empirical findings¹², correspond to: the early terminators (≤4 weeks of use); those who used NRT for the minimum period (5–8 weeks); those who used NRT for the standard recommended period (9–12 weeks); those who extended their use of NRT according to revised recommendations (13–24 weeks)¹⁰; and those who exceeded the maximum period of NRT use (>24 weeks). The study participants were categorized into one of the five mutually exclusive groups according to their longest purchase episode, absent a purchase lapse exceeding 28 days. For example, if a participant had purchased NRT on 5 February, 21 February and 16 March, and made a subsequent purchase on 9 June, then the participant was classified as having purchased NRT for 5–8 weeks of use (range: 29–56 days). For those participants who had multiple but equal purchase episodes, the longest episode in days was selected. An interruption in the purchase of NRT exceeding one month was used previously in signifying a new episode of use⁹. Our classification scheme did not account for those who made bulk purchases of NRT for use over a lengthy time period. Furthermore, the classification scheme was not based on incident purchases, and, thus, does not necessarily represent individuals who used NRT for the first time. For example, an early terminator who made a purchase in January 2016 may have purchased NRT throughout the latter part of 2015.

Based on sales data from the S Group, nicotine gum and lozenge constituted 96.9% of the total sales volume of all nicotine replacement products. Sales of the nicotine patch, which were low and unchanging from 2006 (deregulation) through 2013⁶, were slightly higher in pharmacies compared to other outlets. Given its low overall volume, it is unlikely that the nicotine patch was supplemented with copious purchases of nicotine gum and lozenge. Data released by the S Group included neither the product type (e.g. gum, lozenge, patch) nor product characteristics such as brand, package size, or dosage in milligrams. Expenditures of NRT were available for analysis, but, overlapped too much to differentiate product types. Thus, we opted not to extrapolate package size from expenditure for estimating the supply of NRT for use over a designated period. Instead of extrapolating package size, we used consecutive purchases over 4-week intervals as a measure of duration of NRT use, which does not imply that the purchase(s) provided a sufficient supply of NRT for any given month. Although expenditures of NRT were not used in extrapolating package sizes, they were summed over intervals as a descriptive way of illustrating purchase differences among the five groups of NRT consumers.

Data analysis

The data analyses entailed the use of paired and non–paired statistical tests. Non-parametric methods were employed due to the skewed distributions of purchases and expenditures of NRT and cigarettes. The Kruskal–Wallis test on ranks, analogous to a one-way ANOVA, was used for comparing purchases and expenditures across the five groups of NRT consumers. One notable comparison across groups was the number of NRT purchases made in the first four weeks of a purchase period. The Wilcoxon signed rank test was used for testing differences within pairs across time periods, by NRT consumer group. Purchases in the first four weeks served as the baseline for comparisons to purchases made in subsequent 4-week intervals. The within-pair tests facilitated assessment of whether NRT purchases stayed the same or diminished following the first four weeks of a purchase episode.

Characteristics of NRT consumers

The 728 participants made a total of 9300 NRT and 21601 cigarette purchases in 2016. Most purchases of NRT and cigarettes occurred on Friday, making up 16.2% and 17.4% of total sales, respectively. A higher per cent of females versus males (66.9% vs 57.0%) and young versus the old (72.9% vs 61.6%) prematurely terminated use of NRT. Tests of independence indicated a statistically significant association between NRT consumer group and sex (χ² ₍₄₎=10.8; p=0.03), but not age group (χ² ₍₈₎=10.5; p=0.3) (Table 1).

The early terminators, who did not make consecutive purchases across 4-week intervals, composed almost two-thirds of the entire sample. Of the NRT products, 15.2% and 11.6% were purchased by the early terminators during January 2016 and the other eleven months of the year (χ² ₍₁₎=7.8; p=0.005), respectively, suggesting that several early terminators may have purchased the medication as part of a New Year’s resolution. Among the early terminators (n=462), 237 had purchased NRT only one time with a median expenditure of €6.6 (range: €2 – €30.1). These onetime purchasers of NRT frequently purchased cigarettes as evidenced by a high median number of cigarette purchases (31) and cigarette expenditures (€251.8).

The participants who exceeded NRT recommendations (>24 weeks, n=96), referred to as persistent users, had the fewest cigarette purchases due in part to the large per cent (50%) of users who did not make a single cigarette purchase in 2016. The persistent users had a median purchase gap of NRT that was half the value observed in those who bought NRT over the recommended period (6 vs 12; for all groups, χ² ₍₄₎=82.5; p<0.0001). The purchase gaps were assessed within periods of consecutive NRT purchases (gaps ≤28 days), thus negating the possibility that large gaps could be attributed to a later purchase of NRT. Almost 5% of the entire sample (n=35) had purchased NRT for 13 consecutive 4-week intervals in 2016.

Purchases of NRT over time

For the first month of use, purchases of NRT varied significantly across the five groups of consumers (χ² ₍₄₎=416.9; p<0.0001), with the median purchase ranging from one to four in the early terminators and persistent users, respectively (Table 2). Similar purchase patterns were observed across the NRT consumer groups for the subsequent 4-week intervals. In examining within-pair differences, we observed statistically significant declines in NRT purchases from baseline (i.e. first four weeks of a purchase episode) to the subsequent 4-week intervals in the groups using NRT over the minimum, recommended, and extended time periods. For example, those who followed minimum recommendations had a median of three purchases in the first month of use, but only 1.5 purchases in weeks five through eight (S=624.5; p<0.001). Declines following the first four weeks of purchasing NRT were evident for every group with exception those who exceeded the guidelines. The latter had a median of four purchases in weeks 1–4, weeks 5–8, weeks 9–12, and weeks 13–16. By weeks 25–28, a significant decline in NRT purchases from baseline had been observed (S=760.5; p<0.001). Yet, the persistent users’ purchases and expenditures were rather stable over the whole period.

Concurrent purchases of cigarettes and NRT

The percentage of concurrent purchases of cigarettes and NRT in the first and last month of a purchase episode, shown in Figure 1, excludes the participants who never purchased cigarettes in 2016. Concurrent purchases were coded in terms of pairs of consecutive purchases of the two products. For example, if a customer purchased a cigarette pack on day 5, another cigarette pack on day 17, and NRT on day 20, then a single concurrent purchase would be coded as having occurred within the same week. In this way, the coding scheme selected the shortest period between a cigarette and NRT purchase.

Figure 1

Concurrent purchases of cigarettes and NRT in the first month (n=844 purchases) and last month (n=286 purchases) of a purchase episode

In their single month of NRT use, the early terminators had a lower percentage of concurrent purchases relative to other groups, which occurred despite frequent purchases of cigarettes throughout the year. Concurrent purchases on the same day or same week, which occurred frequently for the other groups, changed slightly among those who purchased NRT over minimum and extended periods. In contrast, the decline in the percentage of concurrent purchases on the same day/week was more evident for those who followed recommendations (75.7% to 56.8%) and those who exceeded recommendations (75.2% to 63.7%).

Limitations

The time-series analysis of retail purchases of cigarettes and NRT minimized the measurement error that frequently occurs in survey data. Yet, there are limitations that need to be acknowledged with regard to using purchase data. First, the cigarette and NRT purchases could have been used by someone other than the purchaser, such as a family member. Data indicating more than one smoker per smoking household⁹ suggest the possibility of multiple quit attempts within the same household. Consequently, single purchasers of NRT within such households could have inflated our estimate of NRT purchases per customer. The second limitation is that a loyalty member may have purchased NRT in an S Group market, but purchased cigarettes elsewhere (or vice versa); consequently, neither smoking status nor prevalence of NRT use could be determined. Third, our categorization of NRT consumer groups was based on purchase episodes in 2016, not purchase incidents. This limitation precluded us from estimating an individual’s risk of becoming a persistent user of NRT. We could not determine if the longest purchase episode of NRT occurred in 2016, or some period prior to 2016. Further, we could not determine if an individual who purchased NRT in the latter part of 2016 continued to purchase NRT throughout 2017.

An additional limitation was the inability to assess whether the study’s findings can be generalized from the sample of loyalty members who purchased NRT to the population of NRT consumers in southern Finland. One surrogate measure for purchasing NRT is smoking status. Yet, the absence of data on smoking status in the sample and population precluded us from weighting our estimates according to population proportions of smokers and non-smokers. Since the highly educated were oversampled in our study, it is likely that any such weighting scheme would have yielded a proportion of early terminators of NRT that exceeded the proportion reported in the current study. Yet, in the absence of a weighting scheme, the estimate for early termination of therapy was already inflated by the higher percentage of female customers (who were oversampled) versus male customers who terminated purchases of NRT prematurely. In contrast, Balmford et al.¹³ reported that female smokers had greater odds, though nonsignificant, than male smokers of completing a course of treatment (OR=1.24, 95% CI: 0.95–1.61).

The absence of product type (e.g. gum, lozenge) and product characteristics (e.g. package size) precluded an accurate assessment of whether a consumer purchased a sufficient amount of NRT to cover a 4-week period. The alternative of using more detailed data from Nielsen Company’s Homescan Consumer Panel was not an option because of the insufficient number of participating households in Finland (n=5000; pers. comm. Ethan Markovitz, Manager, Client Solutions, Nielsen). Even if more detailed data were available, such information would not have accounted for the variability in daily use of oral NRT⁹.