Precise Estimation for the Age of Initiation of Tobacco Use Among U.S. Youth: Finding from the Population Assessment of Tobacco and Health (PATH) Study, 2013‐2017
Folefac D Atem1,2, Meagan A Bluestein1,3, Melissa B Harrell1,3,4 ,Baojiang Chen1,3, Sarah E.Messiah1,2, Arnold E Kuk1,3, Kymberle L Sterling1,3, Charles E Spells1,3 and Adriana Pérez1,3
1The University of Texas Health Science Center at Houston, School of Public Health, USA
2Center for Pediatric Population Health, Children’s Health System of Texas, and UT Health, USA
Science Center School of Public Health, School of Public Health, USA
3Michael & Susan Dell Center for Healthy Living, USA
4Consultant with litigation involving the vaping industry, USA
Submission: October 07, 2022; Published: October 21, 2022
*Corresponding author: Dr. Adriana Pérez, The University of Texas Health Science Center at Houston, School of Public Health, Michael & Susan Dell Center for Healthy Living, Austin, USA
How to cite this article: Folefac D A, Meagan A B, Melissa B H ,Baojiang C, Sarah E.M, et al. Precise Estimation for the Age of Initiation of Tobacco Use Among U.S. Youth: Finding from the Population Assessment of Tobacco and Health (PATH) Study, 2013-2017. Biostat Biom Open Access J. 2022; 11(1): 555801. DOI: 10.19080/BBOAJ.2022.11.555801
Abstract
Context: Youth tobacco use remains a prominent United States public health issue with a high economic and health burden.
Method: We pooled never and ever users at youth’s first wave of PATH participation (waves 1‐3) to estimate age of initiation for hookah, e‐cigarettes, cigarettes, traditional cigars, cigarillos, and smokeless tobacco prospectively (waves 2‐4). Age of initiation of each tobacco product was estimated using weighted interval‐censored survival analyses. Weighted interval censoring Cox‐proportional hazards regression models were used to assess the association of ever use of the TP at the first wave of PATH participation, sex, and race/ethnicity on the age of initiation of ever use of each tobacco product. Sensitivity analyses were performed to understand the impact of the recalled age of initiation for the left‐censored participants by replacing the recalled age of initiation with a uniform “6” years lower bound.
Results: The proportion of those who ever used each tobacco product at the first wave of PATH participation ranged from (N=580,056 of 32,625 206) for traditional cigars to (N=3,430,105 of 32,958,069) for cigarettes. There was a significant increase in ever use of each tobacco product after the age of 14, with e‐cigarettes and cigarettes showing the highest cumulative incidence of initiation by age 21, while smokeless and cigarillos recorded the lowest cumulative incidence by age 21. The adjusted Cox models showed boys initiated at earlier ages for all of these tobacco products except for hookah, which showed no difference. Similarly, apart from ever use of hookah, non‐Hispanic White youth were more likely to initiate each tobacco product at earlier ages compared to Hispanic, non‐Hispanic Black, and non‐Hispanic Other youth.
Conclusion: The increased sample size and the inclusion of ever users yielded greater precision for age of initiation of each tobacco product than analyses limited to never users at the first wave of PATH participation. These analyses can help elucidate population selection criteria for estimating the age of initiation of tobacco products.
Keywords: Hookah, E‐Cigarette; Cigarette; Traditional Cigar; Cigarillo; Smokeless Tobacco; Left Truncated; Right Censored; Survival Analysis
Introduction
Tobacco product use remains high among United States (US) youth resulting in a substantial public health burden. A report from the 2021 National Youth Tobacco Survey (NYTS) showed that about 5.22 million high school students (aged 14‐18) and 1.34 million middle school students (aged 11‐14) reported having ever tried a tobacco product (TP) in the US [1]. This represents 34.0% and 11.3% of high school students and middle school students, respectively[1]. This high prevalence of tobacco use among youth has resulted in addiction, priming for use of other addictive substances, reduced impulse control, deficits in attention, a deficit in cognition, and mood disorder[2]. This has resulted in more research using comprehensive data to understand the age of initiation of TPs in the contemporary TP marketplace [3‐6].
Despite a recent decrease in youth prevalence of daily cigarette smoking since daily cigarette use doubled in the U.S. in 1980 [7, 8], the consumption of alternative TPs has increased, diminishing the public health gains that resulted from declining cigarette consumption [7,9]. The increase in the use of alternative TPs is noticeable, especially among youth [7, 10]. For example, within the past decade, there have been increases in TP use such as hookah among youth, which has tremendously increased the number of TP consumers [7]. The rising popularity of alternative TPs, including hookah, e‐cigarettes, and cigarillos may quickly diminish the perception that TP use is dangerous and may increase consumption of these products [7, 11].
Within the past decade, efforts have been dedicated to understanding TP initiation behaviors across different products including cigarettes, e‐cigarettes, hookah, and cigar products among never‐users [4, 5, 10, 12‐16]. Previous analyses of the Population Assessment of Tobacco and Health (PATH) study have been based on estimating initiation and the age of initiation prospectively among never‐users of each of these TPs at the first wave of PATH participation [3‐6]. Specifically, not included in these previous analyses are participants who are already users of a TP at their first wave of PATH participation, which in statistical terms means those participants are “left‐truncated” [17]. This is a very common design employed by most epidemiological longitudinal studies [18]. However, this design is subject to a reduction in estimation precision and bias due to left truncation [19,20]. The bias is evident when the proportion of TP users is large or if the distribution of the TP users at the first wave of PATH participation is different from the distribution of the participants that are followed longitudinally for initiation of the TP, which is subject to right‐censoring [21, 22]. In an effort to reduce bias and increase precision, in our analyses we have incorporated users of each TP, as well as never‐users of each TP at the first wave of PATH participation. Thus, we included users who reported their recalled age of initiation and never‐users to prospectively estimate the age of initiation of ever use of each TP. In this study, those who recalled initiating a TP at the first wave of PATH participation were considered left‐censored, while those that were never‐users at first wave of PATH participation, age of initiation was estimated prospectively, and those that never initiated the TP at the end of follow‐up are considered right‐censored. Incorporating both the users of each TP at first wave of PATH participation and prospective follow‐up participants (including right‐censored participants) in the analysis can improve the precision and reduce bias in estimating the age of initiation of ever use of these TPs [19,20]. We estimated the age of initiation of ever use of six TPs: hookah, e‐cigarettes, cigarettes, traditional cigars, cigarillos, and smokeless tobacco [23]. Furthermore, we estimated the age at of youth ever use adjusting for ever use of the TP at first wave of PATH participation, sex, and racial/ethnicity.
Methods
Study Sample Participants. The PATH Study is a national longitudinal study of tobacco use in the U.S. and how it affects the health of its people. This study started in 2013 and employed a complex stratified area probability sampling method to generate a nationally representative sample of youth and adults in the U.S. representing approaximately 46,000 participants. [24, 25]. We used youth (aged 12‐17) data from waves 1‐4 : wave 1: 2013‐2014, wave 2: 2014‐2015, wave 3: 2015‐2016, wave 4: 2016‐2017. Participants who were aged 9 to 11 years at wave 1, (i.e., shadow youth) were invited to enter the study once they turned 12 at waves 2 and wave 3. When youth participants turned 18, they were invited to participate in the adult study for waves 2–4, respectively. All participants were followed‐up through wave 4 [25].
Ever use of a TP at first wave of PATH participation
In the first wave of PATH participation, youth were asked if they ever used each of these TPs. In waves 1‐4, PATH asked youth respondents “Have you ever smoked tobacco in a hookah, even one or two puffs?”. Response options were “yes”, and “no” [5, 26]. Less than 1% of the participants with unknown response, were dropped [5, 26]. In waves 1‐4, PATH asked youth respondents“ Have you ever smoked cigarettes, traditional cigar, cigarillo and smokeless, even one or two puffs? . Response options were “yes”, and “no” [26, 27]. Similarly, less than 1% of participants were dropped for not selected either “yes” or “no”. For e‐cigarette in wave 1, PATH asked youth respondents: “Have you ever used an e‐cigarette, such as NJOY, Blu, or smoking Everywhere, even one or two times?”. Response options were “yes”, and “no”. In waves 2‐4, question was modified: “Have you ever used an electronic nicotine product, even one or two times?“. These questions were assumed to measure the same construct across waves. Response options were “yes”, and “no” [4, 26, 27]. Furthermore, missing data, that is those who didn’t select either “yes” or “no” were dropped [4,26,27].
Interval‐Censored Outcome: Ever Age of Initiation of TP
Our outcome is interval‐censored because the exact date of ever use of each TP is unknown for all participants. The restricted‐use PATH data does not provide the date of birth for youth participants but does provide youth age in years at each wave and the number of weeks between waves. For our primary analyses, TP users at the first wave of PATH participation the lower bound was the recalled age of initiation as reported by the participants at their first wave of PATH participation while the upper bound was the age at the first wave of PATH participation. We also performed sensitivity analyses by replacing the recalled age of initiation of users with a six “6” years as the lower bound, that is, six uniform bound was assumed to be the recall age of initiation for all youth, which assumes a minimum age of 6 years old [28‐30]. For the never‐users at the first wave of PATH participation, their lower bound was the age at first wave of PATH participation while their upper bound was estimated by adding participants’ age at their first wave of PATH participation (waves 1‐3) to the number of weeks between relevant subsequent waves (waves 2‐4) [4]. Never‐users at the first wave of PATH participation who did not initiate the TP during the study period were considered right‐censored, that is, their upper bound age of initiation was unbounded [28,29].
'Covariates: Ever use of a TP at the first wave of PATH participation, Sex and Race/Ethnicity.
In the first wave of PATH participation, participants were asked if the used each of the TP (yes/no: cigarettes, e‐cigarettes, hookah, traditional cigars, cigarillos, filtered cigars, and smokeless tobacco). This binary indicator variable for each TP at first wave of PATH participation was a covariate in the analysis of each one of the TP of interest. Sex was classified as female or male. This variable was imputed by PATH using the household information [23]. Race was assessed as non‐Hispanic White, non‐Hispanic Black, Asian, and Other (including multi‐racial). Ethnicity was categorized as either Hispanic or non‐Hispanic. To be comparable to those in prior Surgeon General’s reports, we classified race/ethnicity into four categories: non‐Hispanic White, Hispanic, non‐Hispanic Black, and non‐Hispanic Other (Asian, multi‐race, and other races) [16, 31, 32].
Statistical Analysis
Survival analyses of the age of initiation of each TP were conducted among PATH youth (ages 12‐17) at their first wave of PATH participation, waves 1‐4[26]. These survival analyses were performed using Statistical Analysis System (SAS version 9.4) macros to estimate the overall distribution of the age of initiation of each of the six TPs PATH provided sampling weights, at their first wave of PATH participation, which is the number of individuals in the population each respondent in the sample is representing [23]. In addition, we incorporated the sampling weights and their corresponding 100 balance replicate weights to account for the complex survey design (see details in their methodology) [23]. A Fay’s correction factor of 0.3 was applied to stabilize the estimates [36-40]. This correction factor provides an alternative to variance estimation which can correct for some of the linearization method’s weaknesses by the replicate weights [39‐41]. For each TP, our primary interpretation was based on the model with the recall age of initiation.
Furthermore, we performed weighted interval‐censored Cox proportional hazards regression to estimate the association of ever use of the TP at the first wave of PATH participation, sex, and race/ethnicity, on the age of initiation of each TP [42]. A key assumption of this model is that these explanatory variables act directly on the baseline hazard function, not on the age of initiation and are constant over time [43]. We had less than 1% missing data in the covariates, we analyzed observation with complete information, that is, the model assumed that the missing data in covariates are non‐informative, in other word, the missingness does not depend on the observed and will have little effect on the estimate of hazard ratios [44].
Results
Weighted summary statistics for demographic characteristics for PATH youth (aged 12‐17 at their first wave of PATH participation) are presented in table 1. The sample size for youth with complete data for each of the TPs was as follows: 17745 youth (representing 33 106 488 U.S. youth) for hookah; 17 594 (representing 32 839 325 U.S youth) for e‐cigarettes; 17 667 youth(Representing 32 958 069 U.S youth) for cigarettes; 17 495 youth (representing 32 625 206 U.S. youth) for traditional cigars; 17 437 youth (representing 32 535 569 U.S. youth) for cigarillos; and 17 602 (representing 32 840 714 U.S youth) for smokeless tobacco. Among these youth, the proportions of those who indicated ever use of each TP at their first wave of PATH participation varied, with 5.7% for hookah, 8.1% for a cigarette, 10.4% for e‐cigarettes, 1.8% for traditional cigars, 4.9% for cigarillos, and 3.5% for smokeless tobacco. The average age of participants at their first wave of PATH participation was 13.9 years old across all TPs. The distribution by sex was evenly distributed, with boys accounting for approximately 51%. The distribution by race/ethnicity was evenly distributed across TPs with approximately 54% non‐Hispanic White, 23% Hispanic, 13% non‐Hispanic Black, and 10% non‐Hispanic Other (Table 2) and (Figure 1) (Panel A is based on the recall age and Panel B is based on “6” years uniform lower bound) present estimates for the cumulative incidence of the estimated age of initiation for each TP. We present estimates of the cumulative incidence based on the recalled age of ever use of each TP and the sensitivity analyses based on uniform “6” years lower bound for the left censored sample.
*Hazards are reported as cumulative percentages (i.e., cumulative incidence)
a: 95% CI: Turnbull 95% confidence interval
¥ PATH restricted file received disclosure to publish: March 02, 2020, and July 23, 2020. United States Department of Health and Human Services.
National Institutes of Health. National Institute on Drug Abuse, and United States Department of Health and Human Services. Food and Drug Administration. Center for Tobacco Products. Population Assessment of Tobacco and Health (PATH) Study [United States] Restricted-Use Files. ICPSR36231‐v13. Ann Arbor, MI: Interuniversity Consortium for Political and Social Research [distributor], November 5, 2019. https://doi.org/10.3886/ICPSR36231.v23
The estimates of the cumulative incidence from both recalled age of initiation and the analyses based on six “6” years lower bound for users at first wave of PATH participation are similar in most cases, suggesting the effects of recall bias on users at first wave of PATH participation to estimate age of initiation are minimal for most of the TPs. At age 14, the cumulative incidence of the age of initiation of e‐cigarette was9.1% (7.9– 10.2) representing 2.9 million youth and by age 17: 36.2% (95% CI: 32.5‐39.9%) of youth, representing about 11.9 million youth. The cumulative incidence of the age of initiation of cigarette use was second to e‐cigarettes with 26.9% (95% CI: 23.7‐30.1%) of youth initiating by age 17, representing approximately 8.9 million USA youth. By age 17, 13.6% (95% CI: 12.6 ‐14.6%) of youth initiated ever use of hookah. This increases to 38.6% (95% CI: 29.3‐47.8%) by age 21. The cumulative incidence of the age of initiation of ever use of cigarillos was 15.9% (95% CI: 14.8 – 17.0%) by age 17. Furthermore, by age 21, the cumulative incidence of youth who initiated ever use of cigarillos was 18.4% (95% CI: 15.0 – 21.8%). By age 17, 10.3% (95% CI:9.3 – 11.4%) of youth representing approximately 3.3 million U.S. youth initiated ever use of smokeless tobacco, and this increases to about 6 million by age 21. Cumulative incidence of ever use of traditional cigars by age 17, was 6.2% (95% CI: 4.2 – 8.1%), but by age 21, the cumulative incidence was 25.5% (95% CI: 19.3 – 31.6).
Table 3 presents results from the interval‐censored Cox proportional hazards models assessing the association of ever use of the TP at the first wave of PATH participation, sex, and race/ethnicity, on the age of initiation for each TP ever use. The analyses revealed the hazard ratios of boys relative to girls of initiating ever use of e‐cigarettes, cigarettes, traditional cigars, smokeless and cigarillos at earlier age was 8%, 13%, 161%, 125%, and 18% higher respectively. Hispanic, non‐Hispanic Black, and non‐Hispanic Other had a 20%, 4%, and 20% increase in the hazard ratio for initiating ever use of hookah at earlier ages compared to non‐Hispanic White youth. The hazard ratio of initiating ever use of e‐cigarette at an earlier age was 27% lower in non‐Hispanic Black youth, and 16% lower in non‐Hispanic Other youth compared to non‐Hispanic White youth. For cigarettes, the hazard ratio of initiating ever use of cigarette at an earlier age was 7% lower in Hispanic youth, 25% lower in non‐Hispanic Black youth, and 11% lower in non‐Hispanic Other youth compared to non‐Hispanic White youth. The hazard ratio of initiating ever use of traditional cigars at earlier ages 41% lower in Hispanic youth, 53% lower in non‐Hispanic Black youth, and 32% lower in non‐Hispanic Other youth compared to non‐ Hispanic White youth. Similar results were found in smokeless tobacco, as the hazard ratio of initiating smokeless tobacco at an earlier age was 44% lower in Hispanic youth, 58% lower in non‐Hispanic Black youth, and 34% lower in non‐Hispanic Other youth compared to non‐Hispanic White youth. Those who indicated ever use of a TP (i.e., hookah, e‐cigarettes, cigarettes, traditional cigars, cigarillos, and smokeless tobacco) at their first wave of PATH participation had a higher risk of an earlier age of initiation across each of the TPs of interest. Similar results were found when we replaced the recall age of initiation with a uniform “6” years lower bound.
Discussion
Previous studies have examined cross‐sectional prevalence of the recalled age of TP initiation among users or have examined age of initiation prospectively among those who are never‐users of these TPs when they enter the study as they are followed‐up longitudinally [3‐6, 10]. Prior cross‐sectional studies present only the median or average age of initiation whereas longitudinal studies present limited to prospective follow‐up; no study estimating the age of initiation of TPs have combined left truncated and prospective follow‐up. Moreover, excluding left‐truncated data (i.e., data from users) could result in substantial bias in estimates, including standard errors that are severely underestimated [19]. Noticeably, our results agree with previously published cross‐sectional studies and prospective longitudinal studies that a good proportion of initiation of these tobacco products occurs prior to age 16 [3‐6,10]. This study is the first to combine left‐truncated and prospective follow‐up to estimate the age of initiation of ever use of six TPs. Our results showed a higher cumulative incidence of ever use of these TPs compared to previously published studies that were limited to non‐users at the first wave of PATH participation into the study. For example, comparing previously published results versus ours, the cumulative incidence of initiation by age 18 was 19.7% (95% CI: 3.9 – 25.4%) versus 22.0% (95% CI: 20.5 – 23.3%) for hookah [5]; 41.7% (95% CI: 37.4 – 45.9%) versus 49.9% (95% CI: 46.3 – 53.5%) for e‐cigarettes [4]; 24.3% (95% CI: 20.7– 28.0%) versus 37.1% (95% CI: 34.3 –40.0%) for cigarettes [6]; and 12.8% (95% CI:10.9 –14.6%) versus 18.4% (95% CI: 16.2 –20.5) for traditional cigars [3].
*Non-Hispanic other includes Asian, multi-race, etc.
We found differences when adjusted for sex and race/ethnicity in the age of initiation for different TPs. Our results agreed with a previous cross‐sectional study of hookah use that showed no difference in initiation between boys and girls [10], but not with a prospective longitudinal study that showed girls are more likely to initiate hookah at earlier ages compared to boys [5,10]. For cigarettes and e‐cigarettes, our results for the age of initiation by sex and race/ethnicity resulted in similar interpretations to previously published longitudinal studies [4,6]. Specifically, a previous study of the prospectively estimated age of e‐cigarette initiation in PATH youth (aged 12‐17 at their first wave of participation) showed boys were more likely to initiate e‐cigarettes at earlier ages than girls, and non‐Hispanic White youth were more likely to initiate e‐cigarettes at earlier ages than non‐Hispanic Black and non‐Hispanic Other youth [4]. Similarly, in a different prospective study that examined the age of cigarette initiation among PATH youth (aged 12‐17 at their first wave of participation) never‐users, boys were more likely to initiate cigarettes at earlier ages than girls, and non‐Hispanic White youth were more likely to initiate cigarettes at earlier ages than Hispanic, non‐Hispanic Black, and non‐ Hispanic Other youth [6, 10]. These interpretations are similar to previously published studies on ever use of cigarettes and e‐cigarettes; however, our hazard ratios are slightly higher [4,6,10,45].
Our results for the age of initiation of smokeless tobacco were similar by race/ethnicity to previous published cross‐sectional but different by sex as this study showed no significant difference in initiation by sex [10]. Regarding traditional cigars and cigarillos, our estimates of the cumulative incidence of the age of initiation of these TPs slightly from previously published estimates [3, 10]. There was no significant difference by sex and race/ethnicity from a previously published study [3]. The primary strength of our study is finding that excluding left truncated data among users could introduce bias in estimates and reduction in precision [19]. This is shown in our study in which our confidence intervals are narrower as compared to previous studies that relied on cross‐sectional data or prospective longitudinal samples of never‐users at their first wave of PATH participation, indicating higher precision in our results [3‐6, 10, 45]. As noted above, the inclusion of left‐truncated data increases the estimate of the cumulative incidence of those who initiate TP use, as well. In addition to the primary analyses, we conducted sensitivity analyses for each model using a uniform “6” years lower bound and their age at first wave of PATH participation as the upper bound for this sample which is another strength. The greatest limitation of this study is that for the left‐truncated sample, users might not remember their age of initiation accurately. However, this limitation is inherent in any study of the age of initiation that relies on participant recall.
Conclusion
Our results showed increases in the cumulative incidence of initiation of TPs compared to previously published prospective longitudinal studies (never‐users at their first wave of participation) or cross‐sectional studies of recall (users at the wave of the first wave of participation). Example, for e‐cigarettes and cigarettes since the proportion of youth that indicated ever use of these TPs at their first wave of PATH participation were higher than in other TPs. Therefore, these analyses can help elucidate appropriate population by TPs for more accurately and more precisely estimation of the age of initiation of TPs. In some TPs such as traditional cigars, smokeless tobacco and cigarillos where the proportions of ever use at the first wave of PATH participation were 1.8%, 3.5% and 4.9%, respectively, including those that indicated ever use of a TP might be a waste of time, while in other TPs such as hookah, e‐cigarettes and cigarettes were the these proportions were 5.7%, 8.1% and 10.4% respectively, might be helpful to include those who indicated ever use at the first wave of PATH participation in the analyses. Preventive intervention that incorporates these differences might help in reducing the use of these TPs.
Author Contribution
FDA, AP, MBH, and SEM, wrote the diversity supplement protocol. FDA, BC, MAB, AEK, and AP conducted the searches and screening, extracted, and analyzed the data, and wrote the first draft of the manuscript. MBH, SEM, KLS, CES, AP supervised and redrafted the manuscript. FDA and AP assembled the final draft and circulated it to all the authors who read and agreed on the final submitted draft. AP obtained funding.
Declarations
The institutional review board (IRB) approval for this study was obtained from the Committee for the Protection of Human Subjects at the University of Texas Health Science Center at Houston with the number HSC‐SPH‐17‐0368.
Consent to Participate
PATH obtained parental or legal guardian consent of the youth and youth provided assent. Once youth turned 18 years old, they provided consent again to continue in the PATH study.
Competing Interests
Dr Harrell is a consultant in litigation involving the vaping industry. Other authors declare no competing interests.
Acknowledgement
We thank Dr. Cristine Delnevo for her directions and insightful comments of the manuscript.
References
- Gentzke, Andrea S, Teresa W Wang, Monica Cornelius, Eunice Park‐Lee et al. (2022) Tobacco Product Use and Associated Factors Among Middle and High School Students—National Youth Tobacco Survey, United States, 2021. MMWR Surveillance Summaries 71(5): 1-29.
- Murthy Vivek H (2017) E‐cigarette use among youth and young adults: a major public health concern. JAMA pediatrics 171(3): 209‐210.
- Chen Baojiang, Kymberle L Sterling, Meagan A Bluestein, Arnold E Kuk, Adriana Perez et al. (2020) Age of initiation of cigarillos, filtered cigars and/or traditional cigars among youth: Findings from the Population Assessment of Tobacco and Health (PATH) study, 2013–2017. Plos one 15(12): e0243372.
- Pérez Adriana, Meagan Bluestein, Baojiang Chen, Cheryl L Perry, Melissa B Harrell (2020) Prospectively estimating the age of initiation of e‐cigarettes among US youth: findings from the Population Assessment of Tobacco and Health (PATH) study, 2013‐2017. Journal of biometrics & biostatistics 11(4).
- Perez Adriana, Arnold E Kuk, Meagan A Bluestein, Melissa B Harrell, Cheryl L Perry et al. (2021) Prospective estimation of the age of initiation of hookah use among youth: Findings from the Population Assessment of Tobacco and Health (PATH) study waves 1‐4 (2013‐2017). Addictive Behaviors 117:106838.
- Perez Adriana, Roi San N hpang, Elizabeth Callahan, Meagan Bluestein, Arnold E Kuk et al. (2021) Age at Initiation of Cigarette Use in a Nationally Representative Sample of US Youth, 2013‐2017. JAMA network open 4(2): e210218.
- Minaker Leia M, Alanna Shuh, Robin J Burkhalter, Steve R Manske (2015) Hookah use prevalence, predictors, and perceptions among Canadian youth: findings from the 2012/2013 Youth Smoking Survey. Cancer Causes & Control 26(6): 831‐838.
- McMillen Robert C, Mark A Gottlieb, Regina M Whitmore Shaefer, Jonathan P Winickoff et al. (2014) Trends in electronic cigarette use among US adults: use is increasing in both smokers and nonsmokers. Nicotine & Tobacco Research 17(10): 1195‐1202.
- Connolly Gregory N, Hillel R Alpert (2008) Trends in the use of cigarettes and other tobacco products, 2000‐2007. Jama 299(22): 2629‐2630.
- Sharapova Saida, Carolyn Reyes‐Guzman, Tushar Singh, Elyse Phillips, Kristy L Marynak et al. (2020) Age of tobacco use initiation and association with current use and nicotine dependence among US middle and high school students, 2014–2016. Tobacco Control 29(1): p. 49‐54.
- Roskin Jeremy, Paul Aveyard (2009) Canadian and English students' beliefs about waterpipe smoking: a qualitative study. BMC public health 9(10): 10-12(1-7).
- Apelberg Benjamin J, Catherine G Corey, Allison C Hoffman, Megan J Schroeder, Corinne G Husten et al. (2014) Symptoms of tobacco dependence among middle and high school tobacco users: results from the 2012 National Youth Tobacco Survey. American journal of preventive medicine 47(2): S4‐S14.
- Chen Xinguang, Angela J Jacques‐Tiura (2014) Smoking initiation associated with specific periods in the life course from birth to young adulthood: data from the National Longitudinal Survey of Youth 1997. American Journal of public health 104(2): e119‐e126.
- Dierker Lisa, Joel Swendsen, Jennifer Rose, Jianping He, Kathleen Merikangas et al. (2012) Transitions to regular smoking and nicotine dependence in the Adolescent National Comorbidity Survey (NCS‐A). Annals of Behavioral Medicine 43(3): p. 394‐401.
- DiFranza Joseph R, Judith A Savageau, Kenneth Fletcher, Jennifer O Loughlin, Lori Pbert et al (2007) Symptoms of tobacco dependence after brief intermittent use: the Development and Assessment of Nicotine Dependence in Youth–2 study. Archives of pediatrics & adolescent medicine 161(7):704‐710.
- Wang Teresa W, Andrea S Gentzke, MeLisa R Creamer, Karen A Cullen et al (2019) Tobacco product use and associated factors among middle and high school students—United States. MMWR Surveillance Summaries 68(12): 1-22.
- Kleinbaum David G, Mitchel Klein (2012) Survival analysis: a self‐learning text. Springer 3.
- Mihara Satoko, and Susumu Higuchi (2017) Cross‐sectional and longitudinal epidemiological studies of internet gaming disorder: A systematic review of the literature. Psychiatry and clinical neurosciences 71(7): 425‐444.
- Cain Kevin C, Sioban D Harlow, Roderick J Little, Bin Nan, Matheos Yosef et al. (2011) Bias due to left truncation and left censoring in longitudinal studies of developmental and disease processes. American journal of epidemiology 173(9): 1078‐ 1084.
- Yang Tiantian, Howard E Aldrich (2012) Out of sight but not out of mind: Why failure to account for left truncation biases research on failure rates. Journal of Business Venturing 27(4): 477‐492.
- Matsouaka Roland A, Folefac D Atem (2020) Regression with a right‐censored predictor using inverse probability weighting methods. Statistics in Medicine 39(2): 4001‐4015.
- Qian Jing, Rebecca A Betensky (2014) Assumptions regarding right censoring in the presence of left truncation. Statistics & probability letters 87: 12‐17.
- US Department of Health and Human Services, National Institutes of Health, National Institute on Drug Abuse, US Department of Health, Human Services et al (2020) “Population Assessment of Tobacco and Health (Path) Study [United States] Restricted‐Use Files.” Inter‐university Consortium for Political and Social Research (ICPSR) [distributor].
- Strong David R, Jennifer Pearson, Sarah Ehlke, Thomas Kirchner, David Abrams et al. (2017) Indicators of dependence for different types of tobacco product users: Descriptive findings from Wave 1 (2013–2014) of the Population Assessment of Tobacco and Health (PATH) study. Drug and alcohol dependence 178:257‐266.
- Hyland Andrew, Bridget K Ambrose, Kevin P Conway, Nicolette Borek, Elizabeth Lambert et al. (2017) Design and methods of the Population Assessment of Tobacco and Health (PATH) Study. Tobacco control 26(4): 371‐378.
- US Department of Health and Human Services, National Institutes of Health, National Institute on Drug Abuse, US Department of Health, Human Services Health et al. (2021) Population Assessment of Tobacco and Health (PATH) Study [United States] Restricted‐Use Files. Inter‐university Consortium for Political and Social Research [distributor].
- US Department of Health and Human Services, National Institutes of Health, National Institute on Drug Abuse, US Department of Health, and Human Services et al. (2021) “Population Assessment of Tobacco and Health [United States] Public‐Use Files, User Guide.” Ann Arbor, MI: Inter‐university Consortium for Political and Social Research (ICPSR) [distributor].
- Finkelstein Dianne M (1986) A proportional hazards model for interval‐censored failure time data. Biometrics 42(4): p. 845‐854.
- Finkelstein Dianne M, William B Goggins, David A Schoenfeld (2002) Analysis of failure time data with dependent interval censoring. Biometrics 58(2): 298‐304.
- Zhang Zhigang, Jianguo Sun (2010) Interval censoring. Statistical methods in medical research 19(1): 53‐70.
- Kasza Karin A, Kathryn C Edwards, Zhiqun Tang, Cassandra A Stanton, Eva Sharma et al. (2020) Correlates of tobacco product initiation among youth and adults in the USA: findings from the PATH Study Waves 1–3 (2013–2016). Tobacco control 29(3): s191‐s202.
- Trinidad Dennis R, John P Pierce, James D Sargent, Martha M White, David R Strong et al. (2017) Susceptibility to tobacco product use among youth in wave 1 of the population Assessment of tobacco and health (PATH) study. Preventive medicine 101: 8‐14.
- Klein John P, Melvin L Moeschberger (2006) Survival analysis: techniques for censored and truncated data. Springer Science & Business Media.
- Lindsey J C, L M Ryan (1998) Methods for interval‐censored data. Statistics in medicine 17(2): 219‐238.
- Gentleman Robert, Charles J Geyer (1994) Maximum likelihood for interval censored data: Consistency and computation. Biometrika 81(3):618‐623.
- Goodall Ruth L, David T Dunn, Abdel G Babiker (2004) Interval‐censored survival time data: confidence intervals for the non‐parametric survivor function. Statistics in medicine 23(7): 1131‐1145.
- Ng Meei Pyng (2002) A modification of Peto's nonparametric estimation of survival curves for interval‐censored data. Biometrics 58(2): 439‐442.
- Sun Jianguo (2001) Variance estimation of a survival function for interval‐censored survival data. Statistics in medicine 20(8): 1249‐1257.
- Judkins, David R (1990) Fay’s method for variance estimation. Journal of Official Statistics 6(3): 223‐239.
- Robert E Fay (1989) Theory and application of replicate weighting for variance calculations. In Proceedings of the Survey Research Methods Section of the American Statistical Association.
- Rao JNK, Jun Shao (1999) Modified balanced repeated replication for complex survey data. Biometrika 86(2): 403‐415.
- Betensky Rebecca A, Daniel Rabinowitz, Anastasios A Tsiatis (2001) Computationally simple accelerated failure time regression for interval censored data. Biometrika 88(3): 703‐711.
- David R Cox (1972) Regression models and lifetables. Journal of the Royal Statistical Society: Series B (Methodological) 34(2): 187‐202.
- Carroll Orlagh U, Tim P Morris, Ruth H Keogh (2020) How are missing data in covariates handled in observational time‐to‐event studies in oncology? A systematic review. BMC medical research methodology 20(1):1‐15.
- Edwards Richard, Kristie Carter, Jo Peace, and Tony Blakely (2013) An examination of smoking initiation rates by age: results from a large longitudinal study in New Zealand. Australian and New Zealand journal of public health 37(6): 516‐519.