Association of Smoking Cessation and Survival Among Young Adults With Myocardial Infarction in the Partners YOUNG-MI Registry

This cohort study identifies the prevalence of tobacco use and examines the association of both smoking and smoking cessation with survival in a cohort of adults who experienced a myocardial infarction at a young age.

This supplementary material has been provided by the authors to give readers additional information about their work.

eAppendix 1. Definition of Risk Factors
Diabetes was defined as having a fasting plasma glucose >126 mg/dL, a hemoglobin A1c ≥6.5%, or a documented diagnosis and/or treatment for diabetes. Hypertension was defined as a documented diagnosis and/or treatment of hypertension. Dyslipidemia was defined as having a documented diagnosis and/or treatment of dyslipidemia. Obesity was defined as having a body mass index (BMI) ≥30 kg/m 2 or a documented diagnosis of obesity. Family history of premature CAD was defined as a fatal MI, a non-fatal MI, or a coronary revascularization occurring before 55 years of age in the case of first-degree male family members and before 65 years of age for first-degree female family members.
Alcohol use was defined as having a positive toxicology report for alcohol during the index hospitalization or a documented diagnosis and/or treatment of alcohol use. Illicit substance use was defined as having a positive toxicology report during the index hospitalization for one or more of marijuana, cocaine, heroin, or methamphetamines or a documented diagnosis and/or treatment for use of these substances. Depression, anxiety, and psychotic disorders were defined as having a documented diagnosis and/or treatment for these conditions. In addition, we also calculated the atherosclerotic cardiovascular disease (ASCVD) risk score for each individual whenever possible. 1 To adjust for comorbidities, we calculated the Charlson Comorbidity Index, a method of predicting the risk of mortality based on co-morbidities of patients, for each patient as determined by International Classification of Diseases, 9 th Revision (ICD-9) diagnosis and billing codes associated with the index hospitalization. 2 For patients within the sample who were not coded for MI, we added an additional ICD-9 code of "410" (acute MI) prior to calculating the CCI. Additionally, because lower socioeconomic © 2020 Biery DW. JAMA Network Open.
status is known to be associated with all-cause death, household income for each patient was estimated based on household zip codes in conjunction with the 2015 inflation adjusted median household income data provided by the US Census Bureau. 3 Invasive angiographic findings were ascertained through a review of all available cardiac catheterization images by a single interventional cardiologist blinded to all patient outcomes and risk factors. A segment involvement score (SIS) was employed as a measure of the extent of plaque and was calculated as the total number of coronary artery segments exhibiting plaque, irrespective of the degree of luminal stenosis within each segment. For each patient, the SIS ranged from a minimum of 0 to a maximum of 16.

eAppendix 2. Description of Propensity Score
The independent variables included in the score were age, sex, neighborhood stress score (NSS7) type of insurance, segment involvement score, presence of an ST-elevation MI (STEMI), depression, psychotic disorders, obesity, alcohol use, cocaine use, cardiac catheterization, coronary revascularization, as well as statins, and beta-blockers upon discharge.
Of the 910 patients with a smoking status at one-year post-MI, we were able to calculate a propensity score for 876 (96%). The loss of 34 observations was due solely to the absence angiographic data on those individuals. Due to the relatively small number of missing observations, they were handled by means of simple deletion. Among these, 336 (38%) had ceased smoking tobacco, while the remaining 540 (62%) continued to smoke.
We matched 309 (92%) of those individuals who quit smoking post-MI to an individual who continued to smoke. There was significant overlap in the propensity scores between patients who quit smoking and those who did not. Patients who ceased tobacco smoking post-MI had a mean propensity score of 0.43 (95% CI 0.41 to 0.44), while those who continued to smoke had a mean propensity score of 0.35 (95% CI 0.34 to 0.36).
Balance in the propensity score was assessed by means of the stratification method by means of the PSCORE command. 4 Specifically, the full range of the propensity score was first split into 5 equally spaced intervals. Within each interval, the average propensity score of the treated (i.e. those who quit smoking) was compared to the average propensity score of the control (i.e. those who continued to smoke) to ensure that there was no difference.

eAppendix 4. Sensitivity Analysis
We used two methods to assess the validity of our findings. First, we determined how sensitive the results of the model were to the individual removal of each of the included covariates. The goal of doing so would be to mimic the effect of an influential yet unmeasured confounder on the stability of our model. The relative hazard of all-cause death changed by an absolute maximum of 0.12 when any single variable was removed but continued to be significant in each scenario. The relative hazard of CV death changed by an absolute maximum of 0.14 when any single variable was removed but, as with allcause mortality, remained significant. Second, we repeated the analysis of smoking cessation after weighting by the inverse probability of treatment (IPTW) to account for any possible bias in self-selection for smoking cessation and, hence, a discrepancy between the average treatment effect on the treated (ATT) and the average treatment effect (ATE) of smoking cessation. 5,6 Stabilized weights 6 were used to reduce the influence of outliers on estimates of the effect of smoking cessation on the outcome in question. 6,7 After weighting, the adjusted hazard ratios of smoking cessation on all-cause death and CV death were 0.26 (CI 0.12 to 0.54, p<0.001) and 0.19 (CI 0.05 to 0.69, p=0.01), respectively. Error in the estimation of the weights was assumed to be negligible. Moreover, the robust sandwich-type variance estimator was used as it would provide the most conservative measure of significance. 8 © 2020 Biery DW. JAMA Network Open.

Data
Of the 1053 individuals determined to be current smokers at the time of their MI and who survived for more than one year following their index infarction, smoking status at oneyear post-MI was unavailable for 143. Differences in the baseline characteristics between the those for which data was present (n=910) and those for which it was absent (n=143) are presented in Supplemental Table 2. Those patients for which no follow-up data were available at one-year post-MI were less likely to be diagnosed with psychotic disorders or depression, less likely to be obese, less likely to use alcohol. Despite these differences, there was no significant difference in all-cause mortality, with hazard ratio for no followup of 1.02 (CI 0.60 to 1.74), p=0.95. Similarly, for cardiovascular mortality the hazard ratio for no follow-up was 0.82 (CI 0.38 to 1.78), p=0.62. Similar results were found after including those patients for which we were unable to calculate a propensity score.