Manderski et al. stated that our conclusion that e-cigarette use was associated with having had a myocardial infarction1 was erroneous because we did not use population weights in our analysis. We did not use weights when presenting the descriptive statistics for the sample (Table 1 in our paper1) because we were describing the actual (unweighted) sample characteristics as opposed to presenting population estimates based on the sample (which would have used the weights). The logistic regression analyses of the association between e-cigarette use and having had a myocardial infarction (Table 2 in our paper1) used the National Health Interview Survey (NHIS) weights, accounted for the complex survey design, and followed NHIS procedures to combine the 2014 and 2016 data sets.2 Our multivariable analysis is nationally representative and does not underestimate variance.
Table 1.
Multivariable Associations Between E-Cigarette Use and Myocardial Infarction of NHIS 2014, 2015, and 2016
| Characteristics | 2014 | 2015 | 2016 | Combined NHIS 2014–2016 | ||||
|---|---|---|---|---|---|---|---|---|
| OR (95% CI) | p-value | OR (95% CI) | p-value | OR (95% CI) | p-value | OR (95% CI) | p-value | |
| E-cigarette use | ||||||||
| Never | ref | ref | ref | ref | ||||
| Former | 1.17 (0.84, 1.61) | 0.355 | 1.44 (1.04, 1.98) | 0.028 | 1.02 (0.74, 1.41) | 0.901 | 1.16 (0.96, 1.40) | 0.128 |
| Some days | 1.72 (1.03, 2.87) | 0.037 | 2.02 (1.17, 3.50) | 0.012 | 0.76 (0.42, 1.39) | 0.371 | 1.49 (1.06, 2.09) | 0.022 |
| Daily | 2.06 (0.90, 4.68) | 0.086 | 2.29 (1.26, 4.15) | 0.006 | 2.20 (1.16, 4.17) | 0.016 | 2.14 (1.41, 3.25) | <0.001 |
| Cigarette smoking | ||||||||
| Never | ref | ref | ref | ref | ||||
| Former | 1.83 (1.51, 2.21) | <0.001 | 1.46 (1.20, 1.78) | <0.001 | 1.64 (1.38, 1.95) | <0.001 | 1.64 (1.48, 1.82) | <0.001 |
| Some days | 2.45 (1.56, 3.85) | <0.001 | 1.89 (1.19, 3.01) | 0.007 | 1.93 (1.27, 2.93) | 0.002 | 2.08 (1.60, 2.70) | <0.001 |
| Daily | 3.01 (2.23, 4.06) | <0.001 | 2.16 (1.63, 2.86) | <0.001 | 2.81 (2.15, 3.69) | <0.001 | 2.68 (2.29, 3.13) | <0.001 |
| Hypertension | 2.86 (2.28, 3.59) | <0.001 | 2.05 (1.65, 2.56) | <0.001 | 1.99 (1.62, 2.46) | <0.001 | 2.28 (2.01, 2.58) | <0.001 |
| Diabetes mellitus | 1.77 (1.44, 2.17) | <0.001 | 2.28 (1.87, 2.78) | <0.001 | 1.94 (1.57, 2.40) | <0.001 | 1.99 (1.76, 2.25) | <0.001 |
| High cholesterol | 2.16 (1.77, 2.63) | <0.001 | 2.66 (2.22, 3.19) | <0.001 | 2.65 (2.20, 3.20) | <0.001 | 2.45 (2.20, 2.74) | <0.001 |
| Woman | 0.52 (0.43, 0.62) | <0.001 | 0.50 (0.42, 0.60) | <0.001 | 0.46 (0.38, 0.55) | <0.001 | 0.49 (0.44, 0.55) | <0.001 |
| Age (per 10 years) | 1.61 (1.51, 1.73) | <0.001 | 1.81 (1.69, 1.94) | <0.001 | 1.70 (1.56, 1.86) | <0.001 | 1.70 (1.63, 1.77) | <0.001 |
| BMI | 1.01 (0.99, 1.02) | 0.444 | 1.01 (0.99, 1.02) | 0.330 | 0.99 (0.98, 1.01) | 0.402 | 1.00 (0.99, 1.01) | 0.588 |
| Race/ethnicity | ||||||||
| White | ref | ref | ref | ref | ||||
| Hispanic | 0.68 (0.53, 0.88) | 0.003 | 0.86 (0.65, 1.13) | 0.267 | 0.90 (0.62, 1.29) | 0.553 | 0.81 (0.68, 0.97) | 0.025 |
| Black | 0.95 (0.72, 1.24) | 0.684 | 0.82 (0.64, 1.05) | 0.117 | 0.81 (0.62, 1.06) | 0.123 | 0.86 (0.73, 1.01) | 0.058 |
| Asian | 0.68 (0.43, 1.10) | 0.114 | 0.75 (0.44, 1.29) | 0.298 | 0.38 (0.20, 0.72) | 0.003 | 0.58 (0.43, 0.79) | 0.001 |
| Other race | 1.44 (0.73, 2.82) | 0.294 | 1.35 (0.66, 2.75) | 0.414 | 1.45 (0.64, 3.26) | 0.370 | 1.42 (0.92, 2.18) | 0.115 |
| n | 35,156 | 30,752 | 31,744 | 97,652 | ||||
| Population size (weighted size) | 229,754,972 | 220,309,324 | 235,393,015 | 228,485,770 | ||||
NHIS, National Health Interview Survey.
There are several differences in the way that Manderski and colleagues did their analysis compared with what we did. (At our request, the editor obtained the SAS code so that we could understand precisely what they did.) First, Manderski et al. used the NHIS “Weight -Final Annual” (WTFA) in their analysis. WFTA is based on design and ratio (including nonresponse and post-stratification) adjustments. This weight was designed mainly for the analysis of the family and person data. By contrast, we used “Sample Adult Weight - Final Annual” (WTFA_SA) because it includes design, ratio, nonresponse, and post-stratification adjustments for sample adults. According to the Centers for Disease Control and Prevention, “National estimates of all sample adult variables can be made using these weights.”2 We used WTFA_SA because all variables that we used in our analysis were from the adult sample except for race/ethnicity. Second, Manderski and colleagues used age as a continuous variable in their logistic regression model, and then used the “ESTIMATE” function to measure the effect of a 10-year interval. We used age in 10-year intervals as a variable (i.e., age/10 years); the resulting roundoff errors in the calculations contribute to the small differences between our results. Third, Manderski et al. treated BMI=99.99 as a missing value, whereas we incorrectly treated this as a real value. We updated the results in this letter, treating BMI=99.99 as a missing value. This change did not substantially alter the results. Most importantly, despite the differences in the analytic approach of Manderski and colleagues, they obtained essentially the same results that we did for the combined data for 2014 and 2016, namely a significant association between daily e-cigarette use and myocardial infarction (OR=1.74, 95% CI=1.14, 2.64 in their analysis; OR=1.79, 95% CI=1.20, 2.66 in ours).
We did not use the 2015 data in our paper because we did not realize that e-cigarette use was available, as it was buried in the cancer control supplement. Table 1 in this letter shows that the 2015 data reveal significantly increased odds of having had a myocardial infarction for both some-day and everyday e-cigarette use. The overall risks including 2015 are higher than we originally reported1 based on just 2014 and 2016 (OR=1.49 for all 3 years vs 1.16 for 2014 and 2016, for e-cigarette use on some days; OR=2.14 vs 1.78 for everyday e-cigarette use). Moreover, with the additional 2015 data, some-day e-cigarette use became statistically significant.
Manderski et al. argued that an analysis of e-cigarette use among people who never smoke is necessary to assess an independent association between e-cigarettes and myocardial infarction. As we explained in our response to the first letter written about our paper,3 we do not need to perform this type of analysis to estimate the association between e-cigarette use and having a myocardial infarction because we used multivariable analysis which is adjusted for confounding factors including smoking. Because about two thirds of e-cigarette users also smoke cigarettes (dual users), excluding these people would severely limit the sample size and reduce power to detect a true effect. In addition, because dual use is the dominant pattern for e-cigarette users, it is important to include them in the analysis to obtain the most relevant results.
Similar to the second letter written about our paper, Manderski and colleagues argued that use of the term “risk” implies causation. As we responded before,4 the term “risk factor” was developed to specifically describe cross-sectional studies.
Finally, three recent studies have confirmed our results. The first study used the data of the Behavioral Risk Factor Surveillance System and found elevated risks for myocardial infarction and stroke,5 the second used the data of the Population Assessment of Tobacco and Health study,6 and the third study used NHIS data (2014, 2016, and 2017).7 All three yielded similar risks as we found in the NHIS.
ACKNOWLEDGMENTS
Dr. Glantz’s work was supported by grants R01DA043950 from the National Institute on Drug Abuse, U54HL147127 from the National Heart, Lung, and Blood Institute, and from the Food and Drug Administration (FDA) Center for Tobacco Products. The content is solely the responsibility of the authors and does not necessarily represent the official views of NIH or FDA. The funding agencies played no role in study design; collection, analysis, and interpretation of data; writing the report; or the decision to submit for publication.
No financial disclosures were reported by the authors of this paper.
Contributor Information
Talal Alzahrani, Department of Medicine, The George Washington University, Washington, District of Columbia.
Stanton A. Glantz, Department of Medicine, Cardiovascular Research Institute, Philip R. Lee Institute for Health Policy Studies, Center for Tobacco Control Research and Education, University of California, San Francisco, San Francisco, California.
REFERENCES
- 1.Alzahrani T, Pena I, Temesgen N, Glantz SA. Association between electronic cigarette use and myocardial infarction. Am J Prev Med. 2018;55(4):455–461. 10.1016/j.amepre.2018.05.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.CDC. National Health Interview Survey (NHIS) Public Use Data Release: survey Description (page 99); 2016. ftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/Dataset_Documentation/NHIS/2016/srvydesc.pdf. Published 2017. Accessed September 20, 2017.
- 3.Alzahrani T, Pena I, Temesgen N, Glantz SA. E-cigarettes: stick to the evidence. Am J Prev Med. 2019;56(1):160–161. 10.1016/j.amepre.2018.09.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Alzahrani T, Glantz S. The association between e-cigarette use and myocardial infarction is what one would expect based on the biological and clinical evidence. Am J Prev Med. 2019;56(4):627 10.1016/j.amepre.2018.11.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Ndunda PM, Muutu TM. Abstract 9: electronic cigarette use is associated with a higher risk of stroke. Stroke. 2019;50:A9 10.1161/str.50.suppl_1.9. [DOI] [Google Scholar]
- 6.Bhatta D, Glantz S. Electronic cigarette use and myocardial infarction among adults in the US in the Population Assessment of Tobacco and Health. J Am Heart Assoc. 2019;8(12):e012317 10.1161/JAHA.119.012317. [DOI] [PMC free article] [PubMed] [Google Scholar] [Retracted]
- 7.Vindhyal MR, Ndunda P, Munguti C, Vindhyal S, Okut H. Impact on cardiovascular outcomes among e-cigarette users: a review from National Health Interview Surveys. J Am Coll Cardiol. 2019; 73(9 suppl 2):11 10.1016/S0735-1097(19)33773-8. [DOI] [Google Scholar]