AMI-1

Cardiovascular and mortality risks in older Medicare patients treated with varenicline or bupropion for smoking cessation: an observational cohort study

David J. Graham1*, Kunthel By2, Stephen McKean3, Andrew Mosholder1, Cynthia Kornegay1, Judith A. Racoosin4, Jessica Young3, Mark Levenson2, Thomas E. MaCurdy3,5, Chris Worrall6 and Jeffrey A. Kelman6

ABSTRACT

Purpose To compare cardiovascular and mortality risks in elderly patients treated with varenicline or bupropion for smoking cessation. Methods Elderly Medicare beneficiaries were entered into new-user cohorts of varenicline or bupropion for smoking cessation and followed on therapy for primary outcomes of acute myocardial infarction (AMI), stroke, mortality, and a composite of any of these events. Secondary outcomes were unstable angina, coronary revascularization, and a composite of any primary or secondary outcome event. Propensity score stratification was used to adjust for baseline differences in potential confounding factors. Hazard ratios (HRs) with 95% confidence intervals (CIs) were estimated using Cox proportional hazards, with bupropion as reference.
Results In cohorts of 74824 varenicline and 14133 bupropion users, there were 164 AMI, 96 stroke, 87 death, 317 primary composite, and 814 secondary composite events while on therapy. The HRs (95%CI) were 0.79 (0.50–1.24) for AMI, 1.27 (0.63–2.55) for stroke, 0.58 (0.30–1.13) for death, 0.84 (0.58–1.23) for the primary composite, and 0.92 (0.73–1.14) for the secondary composite. The risk of AMI or the primary composite outcome did not differ in subgroups defined by age, diabetes status, or presence of underlying ischemic heart disease. Only 30% of patients remained on either study drug beyond their first prescription.
Conclusion Cardiovascular and mortality risks were not increased in older patients treated with varenicline compared with bupropion for smoking cessation. A potential increase in the risk of stroke with varenicline could not be excluded. Treatment persistence with either drug was low. Published 2014. This article is a U.S. Government work and is in the public domain in the USA.

key words—varenicline; bupropion; cardiovascular risk; mortality; pharmacoepidemiology

INTRODUCTION

Varenicline, a partial agonist with high affinity for the α4β2 nicotinic acetylcholine receptor, attenuates the symptoms of nicotine withdrawal and is approved for use as a smoking cessation aid.1–3 A typical treatment course lasts 12weeks with an additional 12weeks recommended for patients who successfully stop smoking during the initial course of therapy.4 In preapproval clinical trials, smoking abstinence during weeks 9–12 of treatment was achieved in 44% with varenicline, 30% with bupropion, and 17% with placebo. Longterm (52-week) abstinence was achieved in up to 23% with varenicline, 16% with bupropion, and 10% with placebo.4 The most common side effect was nausea.5 Postmarketing reports of adverse neuropsychiatric effects including depression, violent behavior, and suicidal ideation led to the addition of a box warning to US product labeling in 2008.5 adverse events, based on findings from a manufacturersponsored randomized controlled trial.6 A few weeks later, a meta-analysis of 14 randomized placebocontrolled trials reported that varenicline increased the risk of serious cardiovascular events (odds ratio=1.72 (95% confidence interval (CI) 1.09–2.71)), defined as acute myocardial infarction (AMI), unstable angina, coronary revascularization, coronary artery disease, arrhythmia, transient ischemic attack, stroke, sudden or cardiac death, and congestive heart failure.7 Another meta-analysis of 22 randomized placebo-controlled trials, using the same outcome definition for serious cardiovascular events, reported an odds ratio of 1.41 (95%CI 0.82–2.42) and a risk difference of 0.27% (95%CI 0.10–0.63) for varenicline compared with placebo and concluded that varenicline did not increase cardiovascular risk.8 In December 2012, FDA released the results of a meta-analysis of major adverse cardiovascular events (AMI, stroke, and cardiovascular death) performed by the manufacturer at FDA’s request, showing a hazard ratio (HR) of 1.95 (95%CI 0.79–4.82) for varenicline compared with a placebo.9 Recently, an observational cohort study from Denmark found no increase in cardiovascular risk in varenicline compared with bupropion users.10 The median duration of varenicline use in this study was 4weeks, whereas study follow-up lasted 6months.10 Consequently, most of the follow-up time was actually unexposed, raising a concern that this study’s null result could be an artifact of study design. The question of cardiovascular risk with varenicline is important to resolve because cardiovascular disease is the leading cause of death in smokers, and smoking cessation has been shown to reduce death from AMI.11,12 If varenicline was found to increase the risk of serious cardiovascular events, the benefit–risk basis for its use could be altered.
The purpose of this study was to evaluate whether the risk of major cardiovascular events was increased in older Medicare beneficiaries treated with varenicline compared with bupropion for smoking cessation.

METHODS

Medicare provides health insurance coverage to persons aged 65years and older as well as to persons under the age of 65 who have end-stage kidney disease or are disabled.13,14 This study was restricted to the approximately 21 million beneficiaries aged 65years and older enrolled in fee-for-service Medicare Part A (hospitalization), Part B (outpatient medical care), and Part D (prescription drugs) because claims from these sources were necessary for research purposes. For each beneficiary, we linked claims from all settings of care to create a longitudinal record of their health encounters, diagnoses, and drug prescriptions.
A new user, retrospective cohort design was used to compare patients initiating varenicline with those initiating bupropion for smoking cessation.15 From January 2007 through August 2012, we identified all patients who filled at least one prescription for either varenicline or bupropion. Patients were excluded if they had less than 12months of prior enrollment in Medicare Parts A, B, and D; were under the age of 65years; received prior treatment with a study medication; or were in a hospital, a skilled nursing facility or nursing home, or were receiving hospice care on the date of their cohort-qualifying prescription. Additionally, because bupropion is used primarily as an antidepressant in the USA, we excluded all patients from both cohorts who had a diagnosis of psychiatric illness or who were treated with either medications used for psychiatric illness or psychotherapy during the 12months preceding their cohort-qualifying prescription. To further ensure that patients entering the bupropion cohort were being treated for smoking cessation and not depression, we required that the initial prescription was for the 150-mg strength of a sustained release formulation, with pill count and days’ supply data fields indicating twice a day dosing. This corresponds to the dose and formulation of bupropion approved by the FDA for smoking cessation.
During the 12months preceding cohort entry, data on chronic medical conditions, cardiovascular disease, and cardiovascular risk factors were collected for each patient. We also collected data on prescriptions for medications used for the treatment of these conditions (Table 1). Differencesin baseline characteristics between cohorts were assessed using standardized mean differences, calculated as the difference in means or proportions of a variable divided by the pooled standard deviation.16,17 A standardized mean difference of 0.1 or less indicates a negligible difference in means or proportions between groups.
Propensity score stratification rather than matching was used to adjust for differences in the distribution of baseline characteristics and risk factors.18,19 We did this because there were six times as many varenicline as bupropion patients eligible for study inclusion and propensity score matching that would have led to the exclusion of a substantial number of varenicline-treated patients, thereby compromising generalizability.
Propensity score stratification was performed as follows. For all patients meeting study eligibility criteria, unconditional logistic regression was used to estimate the probability of initiating varenicline therapy given their sociodemographic characteristics, baseline medical comorbidities, and medications used during the preceding year. Patients were then classified into strata determined by quintiles of the propensity score in the varenicline cohort, to produce cohorts with similar baseline characteristics and risk factor distributions within each stratum. Adjusted estimates of cardiovascular and mortality risks were obtained by taking a weighted average of the stratum-specific estimates where the weights equaled the number of varenicline patients in each stratum divided by the total number of varenicline patients.20–22 Because the strata were constructed based on quintiles of the varenicline cohort, the weights were 0.2 for each stratum.
Cohort follow-up began on the date of the first qualifying study drug prescription and continued until the occurrence of a study end point, a gap in continuous study medication coverage exceeding 7days, a prescription fill indicating a switch from varenicline to bupropion or vice versa, completion of 180days of continuous use, admission to a hospital for a nonoutcome reason, admission to a skilled nursing facility or nursing home, transfer to hospice care, or the end of the study period, whichever came first. For continuous study drug exposure, we allowed a gap of up to 7days between successive prescriptions, representing 25% of the typical prescription length for varenicline (28days), because we wanted to minimize the misclassification of unexposed follow-up time as exposed. Also, increasing the gap allowance to 28days did not appreciably increase on-therapy cohort persistence. We censored for nonoutcome hospitalization because we believed that most deaths in such patients were likely because of the nonoutcome reason for which they were admitted, not an acute cardiovascular event. We censored for admission to a skilled nursing facility or nursing home because of concerns about incomplete ascertainment of end points in these settings, and we censored for hospice care because most deaths in these patients were expected and therefore unlikely to be because of an acute cardiovascular event.
The primary end points were AMI, stroke, death, and a composite of the three. Secondary end points included coronary revascularization, unstable angina, and a composite of all primary and secondary events. AMI was defined by the International Classification of Diseases (ICD)-9 code 410 in the first or second position of the hospital discharge diagnoses. In recent studies, code 410 had a positive predictive value (PPV) between 89 and 97% in a variety of administrative claims databases.23–27 Of note, code 410 in the first or second position had a PPV of 94% in a recent study using Medicare data.26 Out-of-hospital death occurring within 1day of an emergency department visit for acute ischemic heart disease was also classified as a fatal AMI. Stroke was identified by ICD-9 hospital discharge diagnosis codes 430, 431, 433.x1, 434.x1, and 436, located in the first position only. When listed as the first discharge diagnosis, these codes have a PPV of 92–100% in administrative claims databases.28–30 Death was ascertained by linkage to the Social Security Master Beneficiary Record database, which provides the date, but not cause, of death and captures over 95% of deaths for persons aged 65years or older in the USA.31 The deaths counted in this study were those associated with inpatient AMI or stroke plus out-of-hospital deaths in community-dwelling elderly. Because our censoring criteria excluded hospitalized noncardiovascular and hospice deaths, the outcome of death in this study was designed to be enriched in deaths as a result of sudden unexpected and cardiovascular causes. Unstable angina was defined as hospitalization with a primary diagnosis of ICD-9 code 411. In a large validation study from Canada, this code had a PPV of 86%foracute coronary syndrome (AMI or unstable angina) and 73% for unstable angina.32 Coronary revascularization was defined as ICD-9, diagnosis-related group, Current Procedural Terminology, and Healthcare Common Procedure Coding System codes for coronary angioplasty, stent placement, or coronary artery bypass, similar to those used by health services researchers.33
Cox regression models were used to compare timeto-event between varenicline and bupropion cohorts, with bupropion serving as reference. Statistical significance was determined using 95%CIs and two-tailed p-values (p≤0.05). Incidence rates and rate differences were estimated using event counts and exposure follow-up time. For the primary end points of AMI and the composite of AMI, stroke, and death, the Cox model was also applied to prespecified subgroups defined by age (65–74; 75+), presence of underlying ischemic heart disease, and diabetes status. For the primary end points of stroke and death, there were insufficient events in each of the propensity score strata to perform these subgroup analyses.
We repeated Cox regression in the following sensitivity analyses: (i) with 30 and 365days of follow-up added to the end of the last study drug prescription; (ii) with a gap allowance of 14days; (iii) no longer censoring for nonoutcome hospitalization; and (iv) using 4:1 (varenicline:bupropion) fixed-ratio propensity score matching. The latter analysis was conducted because it would not require the weighting of strata necessitated by propensity score stratification. The expected trade-off was that a substantial portion of varenicline patients would be excluded because a close propensity score match could not be found, thereby limiting generalizability.
This study was performed as part of the SafeRx Project, a joint initiative of the Centers for Medicare & Medicaid Services and the US FDA. It was approved by the Research in Human Subjects Committee of the FDA’s Center for Drug Evaluation and Research. Analyses were performed using Statistical Analysis System (SAS) v. 9.2 (SAS Institute Inc., Cary, North Carolina).

RESULTS

A total of 74824 varenicline and 14133 bupropion patients were qualified for study entry, contributing 9530 and 2063 person-years of on-drug follow-up time, respectively. Mean duration of use was 46.5 and 53.3days, respectively. Of note, only 30% of varenicline and bupropion patients continued treatment beyond their first prescription. Prior to propensity score stratification, cohorts were similar for most baseline covariates except race/ethnicity, geographic region, year of cohort entry, and history of chronic obstructive pulmonary disease or obesity (Table 1). Each propensity score quintile contained 20% of the varenicline cohort, but bupropion patients were most prevalent in the lowest quintile (43%) and least prevalent in the highest quintile (5%). Within each quintile, varenicline and bupropion cohorts were well balanced with respect to all covariates (data not shown).
During follow-up, there were 164 AMI, 96 stroke, 87 death, 317 primary composite, and 814 secondary composite events. The HRs for all primary and secondary outcomes were statistically indistinguishable from the null, indicating no difference in risk for any outcome in patients treated with varenicline compared withbupropion(Table2).However,thepointestimate for stroke was increased (HR= 1.27; 95%CI 0.63–2.55), and the CI was wide. Analyses based on incidence rate differences also showed no difference in risk for any outcome (Table 3). There were no differences in risk of AMI or the primary composite outcome in subgroups defined by age, underlying ischemic heart disease, or diabetes.
Sensitivity analyses in which 30 and 365days of follow-up were added to the end of each patient’s treatment course and in which the gap allowance between successive prescriptions was increased from 7 to 14days, or where censoring was no longer performed for nonoutcome hospitalization, transfer to nursing home, or hospice care, all yielded results similar to the primary analysis (Table 4). A sensitivity analysis based on the 4:1 fixed-ratio propensity score matching included 46680 (62.4%) varenicline and 11670 (82.6%) bupropion cohort members. Close balance was achieved across all covariates included in the propensity score model. The HRs for AMI and the primary composite were very similar to those from the main analysis (Table 5). The HR for stroke in varenicline-treated patients was more increased than in the main analysis (1.56, 95%CI 0.79–3.08), and the HR for death was statistically significantly reduced (0.51, 95%CI 0.30–0.88). In a post hoc sensitivity analysis of our primary results, we broadened our definition of stroke to include hospital discharge diagnoses in the first or second or any diagnosis position. In these, the HR and upper confidence limit moved closer to the null (0.90 and 95%CI 0.50–1.62 and 0.95 and 95%CI 0.54–1.68, respectively).

DISCUSSION

In a large cohort of older Medicare beneficiaries, the risk of AMI, stroke, coronary revascularization, or death was not increased in patients treated with varenicline compared with bupropion for smoking cessation. However, the HR for stroke with varenicline was nonstatistically significantly increased and was accompanied by wide 95%CIs in both the main analysis and also in a prespecified sensitivity analysis that used 4:1 propensity score matching rather than propensity score stratification. In this setting, we are unable to exclude a possible increase in stroke risk. In the primary analysis, the HR for death with varenicline was nonstatistically significantly reduced, but in the propensity score matching analysis, this reduction in HR was statistically significant. This analysis excluded 38% of varenicline and 17% of bupropion users because propensity score matches could not be found, seriously limiting generalizability. We believe that the most appropriate interpretation of these findings is that there is no difference in mortality risk between varenicline and bupropion for smoking cessation and that the reduced HRestimatewithvarenicline in the sensitivity analysis was possibly because of residual confounding. In the USA, bupropion is predominantly used as an antidepressant, and depression is associated with an increased mortality risk.34,35 If some patients within the bupropion cohort were treated for depression rather than for smoking cessation, residual confounding would be possible, with varenicline appearing to have a relatively lower mortality risk than what was actually the case. Also, most patients in our study stopped therapy after receiving only one prescription and hencewere unlikely to have stopped smoking. Finally, the maximum duration of on-drug follow-up in our study was 6months, far too short to observe a health or mortality benefit from any posited superiority in smoking cessation with varenicline over bupropion, had it actually been achieved.36,37
Table 4. Adjusted hazard ratios (95% confidence intervals) for sensitivity analyses in which the duration of cohort follow-up after stopping smoking cessation therapy was increased from 7 to 30 or 365days, in which the gap allowance between successive prescriptions was increased to 14days, or in which patients were no longer censored for admission to hospital, nursing home, skilled nursing facility, or hospice care First degree composite includes the earliest occurrence of AMI, stroke, or death.
Several study limitations should be noted. Although the use of varenicline was fairly high within the Medicare population, bupropion use for smoking cessation was low because the primary use of bupropion in the USA is as an antidepressant. Despite our efforts to exclude patients treated for depression rather than smoking cessation, it is possible that some patients with underlying depression remained in the bupropion cohort, with the potential for residual confounding as previously discussed. Also, treatment persistence was low in this population, with only 30% of vareniclinetreated and bupropion-treated patients continuing therapy beyond the first prescription. This represents the real-world experience of patients enrolled in Medicare and may reflect the fact that the absolute rate of achieving smoking abstinence was low in preapproval Table 5. Sensitivity analysis based on 4:1 propensity score matching showing event counts with adjusted hazard ratios and 95% confidence intervals for cardiovascular and mortality end points in patients treated with varenicline (n = 46680) compared with bupropion (n= 11670) for smoking cessation First degree composite includes earliest occurrence of AMI, stroke, or death.
clinical trials.4,38 The side-effect profile of these drugs may also have contributed to a marked falloff in treatment persistence given that nausea was reported in 30% of varenicline-treated patients and insomnia in 40% of bupropion-treated patients during preapproval trials for smoking cessation.4,38 Of note, treatment persistence with varenicline was similarly low in recent observational studies from Denmark10 and New Zealand.39 Residual confounding by factors not included in our propensity score model is another potential limitation. We had no data on a number of cardiovascular risk factors including detailed smoking history (duration and intensity), physical activity level, and body mass index. If any of these risk factors were preferentially associated with one of our study exposures, biased estimation of the HR could result. However, it seems unlikely that selective prescribing of varenicline or bupropion would be associated with any of these.
Our study had several important strengths. Outcomes were based on well-validated case definitions. Also, our study was large, with nearly 75000 varenicline initiators, yielding narrow CIs for most outcome measures. We preserved the generalizability of our results to all elderly varenicline users by using propensity score stratification, which facilitated the inclusion of all eligible patients initiating varenicline during our study period. This is an important contrast with the Danish cohort study, where only 30% (17926 of 59790) of eligible varenicline patients were included in the analysis because 1:1 fixed-ratio propensity score matching was employed and a match could not be found for the majority of varenicline patients.10 Our analysis was restricted to follow-up while on therapy, increasing the likelihood that we would detect an increased risk of an acute drug-related event if it was present. This is another important distinction between the current study and the cohort study from Denmark. In that study, the median duration of varenicline use was 28days, yet analysis was based on events occurring over 6months of follow-up, analogous to an “intention to treat” approach.10 The null result reported in that study could have reflected the inclusion of substantial amounts of follow-up time during which there was no actual drug use.
In summary, we found that cardiovascular and mortality risks were not increased by varenicline compared with bupropion when used for smoking cessation. Treatment persistence with either therapy was low.

KEY POINTS

• Meta-analyses of randomized trials have yielded conflicting results regarding cardiovascular risks with varenicline, used for smoking cessation.
• An observational study from Denmark reported no increase in cardiovascular risk with varenicline comparedwithbupropion,anotherdrugusedforsmoking cessation. This study used an intention-to-treat approach that misclassified most study time as exposed when it was nonexposed, and only 30% of eligible varenicline-exposed patients were included in the analysis, compromising generalizability.
• The current study used an on-treatment approach and by using propensity score stratification rather than propensity score matching as in the Danish study, included over 99% of eligible vareniclineexposed patients in the analysis.
• The risks of AMI, stroke, and death were not increased in elderly Medicare patients treated with varenicline compared with bupropion for smoking cessation. However, because of a low number of events, an increased risk of stroke could not be excluded.

REFERENCES

1. Mihalak KB, Carroll FI, Luetje CW. Varenicline is a partial agonist at alpha 4 beta 2 and full agonist at alpha 7 neuronal nicotinic receptors. Mol Pharmacol 2006; 70: 801–805.
2. Foulds J. The neurobiological basis for partial agonist treatment of nicotinedependence: varenicline. Int J Clin Pract 2006; 60: 571–576.
3. Hays JT, Ebbert JO. Varenicline for tobacco dependence. N Engl J Med 2008; 359: 2018–2024.
4. Product labeling for Chantix (varenicline) tablets. Available at: http://www.accessdata.fda.gov/drugsatfda_docs/label/2013/021928s030lbl.pdf [10 June 2013].
5. Cahill K, Stead LF, Lancaster T. Nicotine receptor partial agonists for smokingcessation. Cochrane Database Syst Rev 2012; 4: CD006103. doi:10.1002/ 14651858.CD006103.pub6.
6. FDA Drug Safety Communication. Chantix (varenicline) may increase the risk ofcertain cardiovascular adverse events in patients with cardiovascular disease, June 16, 2011. Available at: http://www.fda.gov/Drugs/DrugSafety/ ucm259161.htm [10 June 2013].
7. Singh S, Loke YK, Spangler JG, Furberg CD. Risk of serious adverse cardiovascular events associated with varenicline: a systematic review and meta-analysis. CMAJ 2011. doi:10.1503/cmaj.110218.
8. Prochaska JJ, Hilton JF. Risk of cardiovascular serious adverse events associatedwith varenicline use for tobacco cessation: systematic review and meta-analysis. BMJ 2012; 344: e2856.
9. FDA Drug Safety Communication. Safety review update of Chantix (varenicline) and risk of cardiovascular adverse events, December 12, 2012. Available at: http://www.fda.gov/Drugs/DrugSafety/ucm330367.htm [10 June 2013].
10. Svanström H, Pasternak B, Hviid A. Use of varenicline for smoking cessationand risk of serious cardiovascular events: nationwide cohort study. BMJ 2012: e7176. doi:10.1136/bmj.e7176.
11. Centers for Disease Control and Prevention. Smoking-attributable mortality,years of potential life lost, and productivity losses—United States, 2000-2004. MMWR Morb Mortal Wkly Rep 2008; 57: 1226–1228.
12. Critchley JA, Capewell S. Mortality risk reduction associated with smoking cessation in patients with coronary heart disease: a systematic review. JAMA 2003; 290: 86–97.
13. Centers for Medicare and Medicaid Services. Medicare program general information. Available at: http://www.cms.hhs.gov/MedicareGenInfo/ [26 June 2012].
14. Centers for Medicare and Medicaid Services. Prescription drug coverage-generalinformation. Available at: http://www.cms.hhs.gov/PrescriptionDrugCovGenIn/ [26 June 2012].
15. Ray WA. Evaluating medication effects outside of clinical trials: new-userdesigns. Am J Epidemiol 2003; 158: 915–920.
16. Mamdani M, Sykora K, Li P, et al. Reader’s guide to critical appraisal of cohort studies: 2. Assessing potential for confounding. BMJ 2005; 330(7497): 960–962.
17. Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med 2009; 28: 3083–3107.
18. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika 1983; 79: 516–524.
19. Rosenbaum PR, Rubin DB. Reducing bias in observational studies usingsubclassification on the propensity score. J Am Stat Assoc 1984; 79: 516–524.
20. Greenland S, Robins JM, Pearl J. Confounding and collapsibility in causal inference. Stat Sci 1999; 14: 29–46.
21. Sato T, Matsuyama Y. Marginal structural models as a tool for standardization.Epidemiology 2003; 14: 680–686.
22. Stuart EA. Matching methods for causal inference: a review and a look forward.Stat Sci 2010; 25(1): 1–21.
23. Petersen LA, Wright S, Normand ST, Daley J. Positive predictive value of thediagnosis of acute myocardial infarction in an administrative database. J Gen Intern Med 1999; 14(9): 555–558.
24. Levy AR, Tamblyn RM, Fitchett D, McLeod PJ, Hanley JA. Coding accuracy ofhospital discharge data for elderly survivors of myocardial infarction. Can J Cardiol 1999; 15(11): 1277–1282.
25. Austin PC, Daly PA, Tu JV. A multicenter study of the coding accuracy ofhospital discharge administrative data for patients admitted to cardiac care units in Ontario. Am Heart J 2002; 144(2): 290–296.
26. Kiyota Y, Schneeweiss S, Glynn RJ, Cannuscio CC, Avorn J, Solomon DH.Accuracy of Medicare claims-based diagnosis of acute myocardial infarction: estimating positive predictive value on the basis of review of hospital records. Am Heart J 2004; 148(1): 99–104.
27. Choma NN, Griffin MR, Huang RL, et al. An algorithm to identify incident myocardial infarction using Medicaid data. Pharmacoepidemiol Drug Saf 2009; 18(11): 1064–1071.
28. Tirschwell DL, Longstreth WT. Validating administrative data in stroke research.Stroke 2002; 33(10): 2465–2470.
29. Roumie CL, Mitchel E, Gideon PS, Varas-Lorenzo C, Castellsague J, Griffin MR. Validation of ICD-9 codes with a high positive predictive value for incident stroke resulting in hospitalization using Medicaid health data. Pharmacoepidemiol Drug Saf 2008; 17(1): 20–26.
30. Kokotaila RA, Hill MD. Coding of stroke and stroke risk factors usingInternational Classification of Diseases, revision 9 and 10. Stroke 2005; 36(8): 1776–1781.
31. Hill ME, Rosenwaike I. The Social Security AMI-1 Administration’s Death Master File: the completeness of death reporting at older ages. Soc Secur Bull 2001/ 2002; 64(1): 45–51. Available at: http://www.socialsecurity.gov/policy/docs/ ssb/v64n1/v64n1p45.pdf [26 June 2012].
32. Varas-Lorenzo C, Castellsague J, Stang MR, Tomas L,Aguado J, Perez-GutthannS.Positive predictive value of ICD-9 codes 410 and 411 in the identification of cases of acute coronary syndromes in the Saskatchewan Hospital Automated Database. Pharmacoepidemiol Drug Saf 2008; 17: 842–852.
33. Epstein AJ, Polsky D, Yang F, Yang L, Groeneveld PW. Coronary revascularization trends in the United States, 2001-2008. JAMA 2011; 305: 1769–1776.
34. Schulz R, Beach SR, Ives DG, Martire LM, Ariyo AA, Kop WJ. Associationbetween depression and mortality in older adults. The Cardiovascular Health Study. Arch Intern Med 2000; 160: 1761–1768.
35. Wulsin LR, Evans JC, Vasan RS, Murabito JM, Kelly-Hayes M, Benjamin EJ.Depressive symptoms, coronary heart disease, and overall mortality in the Framingham Heart Study. Psychosom Med 2005; 67: 697–702.
36. Rea TD, Heckbert SR, Kaplan RC, Smith NL, Lemaitre RN, Psaty BM. Smokingstatus and risk for recurrent coronary events after myocardial infarction. Ann Intern Med 2002; 137: 494–500.
37. Teo KK, Ounpuu S, Hawken S, et al. Tobacco use and risk of myocardial infarction in 52 countries in the INTERHEART study: a case-control study. Lancet 2006; 368: 647–658.
38. Product labeling fro Zyban (bupropion) tablets. Available at: http://www. accessdata.fda.gov/drugsatfda_docs/label/2012/020711s036lbl.pdf [30 May 2013].
39. Harrison-Woolrych M, Ashton J. Utilization of the smoking cessation medicinevarenicline: an intensive post-marketing study in New Zealand. Pharmacoepidemiol Drug Saf 2010; 19: 949–953.