The influence of snuff and smoking on bone accretion in late adolescence. The Tromsø study, Fit Futures

The study procedures of The Tromsø study, Fit Futures are published previously [22]. Briefly, the Fit Futures study is a school-based cohort initiated in 2010–2011 (TFF1). All first year upper-secondary school students in Tromsø and the neighbor municipalities were invited to a comprehensive health survey. Out of 1117 invited individuals, 1038 adolescents (508 girls and 530 boys) attended, giving a participation rate of 92.9%. In the follow-up survey, Fit Futures 2 (TFF2) 2 years later (2012–2013), all participants in TFF1 and all new students in third year of the same upper-secondary schools were invited. A total of 66% of the TFF1 cohort met in TFF2 providing 688 repeated measures of aBMD. Participants above 17 years of age at baseline were excluded and 630 individuals 15 to 17 years of age, 349 girls and 281 boys, completed the questions on use of tobacco at both surveys.

Recruitment of participants to both surveys was conducted in close collaboration with the schools. The Clinical Research Unit at the University Hospital of North Norway conducted both health examinations during school days. All participants gave written informed consent at the study site. Participants younger than 16 years of age had to bring a written consent from their superiors to attend the survey. The data collection in TFF1 and TFF2 was approved by the Norwegian Data Protection Authority and the Regional Committee of Medical Research Ethics (REK nord) with a project-specific approval for the present study (Ref. 2019/31193/REK nord).

Changes in bone mineral status were measured by dual-energy X-ray absorptiometry (DXA) as femoral neck (FN), total hip (TH), and total body (TB) bone mineral content (BMC; g) and aBMD (g/cm²). The same instrument (GE Lunar prodigy, Lunar Corporation, Madison, WI, USA) and analytic program (Encore pediatric software [23]) were used in both TFF1 and TFF2. We used auto-analysis software and default region of interest, according to a standardized protocol. Previously, the coefficients of variation ((SD/mean] × 100) for the DXA device used have been estimated to 1.14% at the TH and 1.72% at the FN measured in vivo [24]. We used measurements of the left-sided hip, but in 15 cases, the data was erroneous or missing, and values of the right hip were reported for both TFF1 and TFF2. The main outcome of this study was ΔaBMD; however, ΔBMC is frequently reported to support the understanding of bone accretion.

We collected data on the use of tobacco by electronic self-administered questionnaires. At TFF1, the question “Do you use snuff?” had three alternatives: “no never,” “sometimes,” or “daily.” If answers were “sometimes” or “daily,” participants were asked additional questions on frequency. The questions were as follows: “If you use snuff sometimes, how many snuff portions do you usually take per week?” Alternatives were “one or less,” “2–3,” “4–6,” “7–10,” and “more than 10.” For daily users, the subsequent question was as follows: “If you use snuff daily, how many snuff portions do you usually take per day?” Alternatives were “1,” “2–3,” “4–6,” “7–10,” and “more than 10.” Information about the age of onset were elicited by a question at TFF2: How old were you when you started to use snuff? The 8 alternatives were as follows: “Below 12 years,” “12 years,” “13 years,” “14 years,” “15 years,” “16 years,””17 years,” “18 years,” and “19 years or above.” Questions on smoking had an identical structure as those on use of snuff at both surveys, and only “portions” were replaced by “cigarettes.”

We measured height and body weight to the nearest 0.1 cm and 0.1 kg (Jenix DS 102 Stadiometer, Dong Sahn Jenix, Korea), following standardized procedures with no shoes and light clothing. Based on these parameters, body mass index (BMI) was calculated (kg/m²). Through clinical interviews, we assessed ethnicity (White, Asian, or Black, dichotomized into white or others), the possibility of pregnancy (exclusion criterion for DXA), acute and chronic diseases, use of medication, and use of hormonal contraceptives.

Pubertal maturation information, physical activity level, and alcohol consumption were elicited by the self-administered questionnaire at TFF1. Frequency of alcohol consumption was assessed with a scale from 1 to 5: “never,” “once per month or less,” “2–4 times per month,” “2–3 times per week,” and “4 or more times per week.” We dichotomized the answers into no (never) and yes. Covariates of pubertal maturation in boys were based on the Pubertal Developmental Scale (PDS). Secondary pubertal characteristics as growth spurt, pubic hair growth, changes in voice, and facial hair growth rated on a scale from 1 (have not begun) to 4 (completed) were summarized and divided by four [25]. In girls, pubertal maturation was determined based on self-reported age at menarche.

Self-reported physical activity level was assessed by questions from the modernized Saltin-Grimby Physical Activity Level Scale (SGPALS) [26]. The participants graded their leisure time physical activity in an average week during the last year with four alternatives: “sedentary activities only”; “moderate activity like walking, cycling, or exercise at least 4 h per week”; “participation in recreational sports at least 4 h per week”; “participation in hard training/sports competitions several times a week.” If the activity varied much, for example, between summer and winter, they were asked to give an average.

Hormonal contraceptive use (girls) was categorized into “no,” “estrogen and progestin,” and “progestin only.” We dichotomized answers on use of medication known to affect bone and diseases known to affect bone into yes and no (medication and disease definition, see Table 1).

All statistical analyses were conducted stratified by sex, and characteristics of the study population are presented as means and standard deviations (SD), or count, and percentages. We explored differences by ANOVA with Bonferroni correction and Pearson’s chi-squared test. We calculated absolute change (TFF2 − TFF1) and percentage change ((TFF2 − TFF1)/TFF1*100) in bone traits. Through DXA measurement dates, we were able to compute exact time of follow-up to compute annual change of anthropometric and bone parameters used in crude analyses. For simplicity purposes, the snuff and cigarette frequency answers were categorized into three groups: “ < 1,” “2–6,” and “ > 7” units per week/day.

The associations between the exposure of tobacco and the outcomes of ΔaBMD between TFF1 and TFF2 where investigated by linear regression models. We used TFF2 score as outcome and included the TFF1 score as a covariate to estimate the predictive value of exposure on change (Y2 = β0 + β1Y1 + β2X_snuff + β3…). We compared the results with change-score analysis (Y2-Y1 = β 0 + β1X_snuff) and explored consistency as baseline adjustments in change-score analysis may introduce bias in regression models comparing naturally occurring groups [27].

Initially we conducted crude models. Then potential confounders were added in the following way: model 2, the “anthropometry model,” comprised the crude model plus age, baseline anthropometry (height and weight), and change in anthropometric parameters. In model 3, the full model, pubertal maturation, and baseline physical activity level were added to the “anthropometry model.” In addition, baseline variables previously known to be of clinical importance like ethnicity, alcohol consumption, diagnosis known to affect bone, medication known to affect bone (see Table 1), and hormonal contraceptives use (girls) were then added as covariates using a backwards elimination strategy where p = 0.10 were used as cut-off to stay or leave the model. Based on these elimination procedures, ethnicity was added to all models in boys, and the TH ΔaBMD was adjusted for medication known to affect bone. In girls, hormonal contraceptives use was added to all models. All models were adjusted for time between measurements. Use of snuff models were controlled for daily smoking and vice versa.

During the first few weeks of TFF1, the questionnaire did not contain the questions related to PDS score, giving a high percentage of missing puberty values among boys: n = 52 (18.5%). In girls, six had missing information on menarche age. Multiple imputations based on predictors and outcome variables used in the full model were conducted to predict missing values. We assumed missing at random, 20 imputations were performed, and we report pooled estimates [28]. Normal distribution, linearity, homogeneity, and outliers were explored by residual analysis. In both girls and boys, one outlier was excluded in all models. In girls, the full TB BMC model residuals showed a heteroscedastic pattern, and weighted least square regression was applied. We used menarche age and PDS scores as continuous variables in multiple regression models. Plausible 2-way interactions related to aBMD, age, and pubertal maturation were checked, and interaction terms for age*snuff were added to boys ΔaBMD TB full model and age* “double use” to the ΔaBMD TB full model in girls. Significance level was set to p = 0.05, and all procedures were performed in IBM SPSS Statistics for Windows, version 26 (IBM Corp., Armonk, N.Y., USA).

In girls, 244 (69.9%) were classified as “non-users,” 51 (14.6%) as “sometimes,” and 54 (15.5%) as “daily users” of snuff at baseline. In boys, the corresponding numbers were 185 (65.9%), 31 (11.0%), and 65 (23.1%), respectively. The age of onset was mainly between 13 and 16 years of age, and mean age was 14.6 in girls (n = 93, 12 missing) and 14.7 years in boys (n = 84, 12 missing; 3 participants responding onset under 12 years had age set to 11 years). The age of onset was not correlated with any bone outcomes.

All but six of the “daily users” at TFF1 remained “daily users” during follow-up (51 girls and 62 boys at TFF2). The prevalence of snuff users increased during follow-up. In both girls and boys, roughly 18% “of the non-users” reported sometimes or daily use of snuff 2 years later. In the “sometimes” use of snuff category at baseline, more than half of the girls reported to usually take one portion or less per week and 86.3% reported six portions per week or less. In boys, we observed corresponding numbers, with 58.1% and 74.2%, respectively. Furthermore, we observed that individuals in the “sometimes” category at baseline fluctuated during follow-up, and only 15 of the 82 remained in their initial “sometimes” category. In main analysis, we compared the “non-users” with the “users” of snuff, combining “sometimes” and “daily” users of snuff at baseline. Sensitivity analysis was then conducted with the “sometimes” group excluded, comparing “non-users” with “daily users” group only.

In both girls and boys, the snuff “users” group differed significantly from the “non-users” with a higher prevalence of smokers and alcohol consumers (p < 0.001). Among girls in the “users” category, fewer reported to be engaged in sports activities (p < 0.001), and use of hormonal contraceptives was more prevalent in the “users” group (p < 0.001). In boys, there was a statistically significant difference between the compared groups in annual height (p < 0.001) and weight change (p = 0.020). No differences in baseline aBMD or BMC between the groups were observed.

Crude percentage of bone accretion is shown in Fig. 1. In both girls and boys reporting no use of snuff at baseline, mean annual ΔaBMD was higher compared to change among “users.”

The results of crude and adjusted regression models of ΔaBMD and ΔBMC in relation to use of snuff are presented in Table 2. In girls, the “users” of snuff group had significantly less ΔaBMD compared to the “non-user” group in crude models at both femoral sites (FN: β =  − 0.004, p = 0.028, and TH: β =  − 0.019, p = 0.020), but not at TB. No associations were significant in the fully adjusted models.

In boys, statistically significant associations were observed in both ΔaBMD and ΔBMC, except in the adjusted ΔBMC TB models. Estimated ΔaBMD between “non-users” and “users” of snuff at the FN was 0.012 and 0.015 g/cm² at the TH in the full models, a difference comprising roughly 1% change during follow-up. In anthropometry models, particularly changes in anthropometric measures attenuated the associations.

In analysis where the “sometimes” group was excluded, comparing “non-users” (girls, n = 244; boys, n = 185) with “daily users” (girls, n = 54; boys, n = 65) of snuff showed estimates with negative associations between snuff use and bone accrual. In girls, the “daily users” of snuff group had significantly less ΔaBMD at FN compared to the “non-user” group in adjusted models: β =  − 0.012, p = 0.037, while the TH association also strengthened (β =  − 0.009, p = 0.071). In boys, both ΔaBMD and ΔBMC crude models were statistically significant. However, in partly adjusted models, all associations were attenuated and turned out insignificant. In the fully adjusted models, use of snuff was not statistically significantly associated with ΔaBMD.

In sensitivity analysis related to multiple imputation and the high percentage of missing puberty data in boys, the full models with the original sample (“non-users” vs “users” with original PDS score, n = 229) showed similar estimates, but FN (β =  − 0.015, p = 0.059) and TB (β =  − 0.008, p = 0.062) turned out insignificant.

There were a limited number of daily smokers at baseline in the study population, eight girls (2.3%) and eight boys (2.8%). In the “sometimes” category consisting of 56 girls (16%) and 44 (15.7%) boys, 76.8% of the girls and 47.7% of the boys reported to smoke one or less cigarette a week. In order to obtain a larger comparison group and to enhance exposure of smoking in the “smokers” category, the “one or less” responders were regarded as “non-smokers,” and a combined “daily/more than 2 cigarettes weekly” category consisting of 21 girls and 31 boys was created.

No statistically significant differences in baseline bone traits between smokers and “non-smokers” were observed. Otherwise, the groups differed with a higher prevalence of snuff use, alcohol consumption, and lower physical activity levels in the smokers group compared to the “non-smokers” group. Crude comparisons of change in ΔaBMD are shown in Fig. 2.

The results of crude and adjusted regression models of ΔaBMD and ΔBMC in relation to smoking are presented in Table 3. In girls, no associations between smoking and ΔaBMD were observed, but accumulation of BMC at the FN appeared to be reduced (β = 0.109, p = 0.006).

In boys, the full ΔaBMD TB model was statistically significant (β =  − 0.011, p = 0.037). The association for TH in the crude and anthropometric models was statistically significant but was attenuated in the full model with a borderline association (p = 0.052).

At baseline, 46 girls and 44 boys responded that they were “double users,” indicating responses of either “daily” or “sometimes,” for both smoking and use of snuff at baseline. Among the “daily users” of snuff, 28 (51.9%) of the girls and 32 (49.2%) of the boys reported to be “double users,” i.e., both smokers and daily use of snuff. These individuals were mostly sometimes smokers as only eight girls and eight boys reported smoking daily. Only one out of the 16 daily smokers at baseline was not “double user.”

The baseline differences between “double users” and “non-users” were similar to use of snuff, except that annual change in body weight did not differ in boys. The results of crude and adjusted regression models of ΔaBMD, ΔBMC, and “double use” are presented in Table 4.

In girls, no relationship was observed between ΔaBMD and “double use,” while the ΔBMC was significantly reduced at the FN (p = 0.018). In boys, most models turned out statistically significant, except the adjusted ΔaBMD FN models. An estimated difference in ΔaBMD between “non-users” and “double users” at the TH of 0.018 g/cm² corresponds to ~ 1.6% change during follow-up.

Potential pathophysiologic mechanisms of the adverse effects of tobacco on bone remain to be clarified [38]. Snuff may have different effects on bone than smoking does, because it does not undergo combustion. Nevertheless, the hypothesized influence of tobacco on the skeleton may be indirect through body weight, reduced blood supply, calciotropic hormones, and abnormal PTH/vitamin D axis or direct through nicotine [38‐40]. Studies suggest that snuff generates the same amount of blood plasma nicotine level as smoking [41]. There is a faster uptake of nicotine by smoking, but the blood plasma nicotine level remains higher over a longer period of time by use of snuff [42]. However, the influence of nicotine on bone may be different in growing and mature skeletons [43].

The main strengths of this study are the population-based design and a relatively large, well-described, study population where both sexes are represented. The Tromsø study, Fit Futures is one of few studies investigating adolescent’s health and lifestyle. Trained personnel at the research unit at the University Hospital of North Norway conducted the data collection in order to minimize measurement error. However, the study has some limitations to consider. Only two measurement points and the relatively short follow-up time made it challenging to differ true change in aBMD from measurement error and regression to the mean may be an issue. Even though the precision of DXA is good at a group level, the mean changes of less than 1% are debatable given the CV of the DXA scanner. DXA has limitations because of the two-dimensional measurement of BMD, leading to overestimation of larger bones. The assessment of tobacco exposure by self-administered questionnaire may induce social desirability bias and underreporting of exposure [44]. The lack of information on nutrition and vitamin D levels and the missing PDS scores for 18.5% of males are considered a limitation to the study. Loss to follow-up bias may influence the validity of this study, as the study sample comprises roughly 66% of the original cohort. However, a high attendance rate at baseline (93%) contributed to the information of the drop-out analysis. Proportions of smokers and users of snuff were higher in drop-outs, which could lead to an underestimation of the associations between tobacco and bone accrual in the study. The use of baseline adjustments in two-wave observational studies with naturally occurring groups is debated. Baseline adjustments combined with measurement error may lead to directional bias leaving hypothesis tests vulnerable to type 1 error [45], and comparison with simple change scores without baseline adjustments was conducted according to advices by van Breukelen [27]. When the two approaches do not agree, results should be interpreted with caution. The disagreement was limited to 1 out of 36 models (Supplemental table). Nevertheless, disagreement may also partly be explained by the fact that one approach estimates the total effect of exposure on bone accrual, while the other estimates the direct effect, adjusting for the initial bone trait level [46].