The causal association between smoking and atherosclerotic cardiovascular diseases, malignancies and parenchymal pulmonary diseases has been documented.¹ Despite this clear association, the smoker’s paradox (a better outcome after some particular life threatening diseases in smokers versus non-smokers) has been a matter of debate for many years. Despite the increasing evidence supporting there is not a real smoker’s paradox, there are up-to-date publications claiming there is a smoker’s paradox for some kind of diseases or patient groups, like those with acute myocardial infarction, stroke, and patients undergoing transcatheter aortic valve implantation.^{2 - 4} Low platelet reactivity, enhanced antiplatelet activity with clopidogrel and possible activation of the preconditioning pathways with smoking-induced hypoxia are being suggested as the background of the smoker’s paradox.^{5 , 6} Additionally, it is suggested that some regional disparities can be found for smoker’s paradox.^{7 , 8}

In this study, we aimed to explore the effect of smoking with a dose-dependent approach on very long-term mortality in the survivors of first STEMI.

Three hundred and thirteen consecutive patients diagnosed with acute or subacute STEMI were enrolled in the study. Median follow-up duration was 14.5 years. One hundred and twelve (35.8%) patients had never smoked before the index event (Group 1). The exposure amount of the smokers at the index event was as following; 66 patients (21.1%) <20 package*years (Group 2); 94 patients (30.0%) 20-40 package*years (Group 3); 41 patients (13.1) >40 package*years (Group 4).

Baseline demographic features and laboratory parameters of the smokers and non-smokers are shown in Table 1 . Smokers were younger and more frequently male. Hypertension and diabetes were more prevalent in non-smokers, and family history for coronary artery disease were more prevalent in smokers. While hemoglobin levels at admission were higher among smokers than non-smokers, total cholesterol, low density cholesterol and blood glucose levels were higher at admission in non-smokers than smokers. Infarction site, receiving an urgent reperfusion treatment, peak creatinine kinase (CK) and CK-MB levels, infarct-related artery (IRA) and Gensini scores were similar among smokers and non-smokers ( Table 2 ). Mean ejection fraction (EF) was higher in smokers than non-smokers.

Table 1

Variables	Non-Smokers (112)	Smokers (201)	p
Age (years)	62±11	54±9	<0.001
Gender (Female)	30(26.8)	16(8.0)	<0.001
Hypertension	52(46.4)	46(22.9)	<0.001
Diabetes	28(25.0)	18(9.0)	<0.001
Family History for CAD	17(15.2)	50(24.9)	0.045
Tobacco Exposure - Grade
<20 pack-years	-	66(32.8)
20-40 pack-years	-	94(46.8)
>40 pack-years	-	41(20.4)
Hemoglobin, g/dl	14.0±1.7	14.7±1.5	<0.001
WBC*103	10.8±3.3	11.8±4.0	0.022
Creatinine, mg/dl	1.08±0.26	1.03±0.22	0.120
Total cholesterol, mg/dl	198±44	187±41	0.025
LDL, mg/dl	131±37	121±36	0.017
HDL, mg/dl	41±10	40±11	0.521
Triglyceride, mg/dl	111(68-161)	113(82-164)	0.296
Blood Glucose at admission, mg/dl	148±74	125±45	0.002

CAD: coronary artery disease, HDL: high density lipoprotein, LDL: low density lipoprotein, WBC: white blood cell count Continuous variables were presented as mean±SD or median (IQR 25-75); categorical variables were presented as number (%).

Table 2

Variables	Non-Smokers (112)	Smokers (201)	p
Anterior MI, n(%)	65(58.0)	112(55.7)	0.692
Thrombolysis + Primary PCI, n(%)	90(80.4)	154(76.6)	0.444
Peak CK, U/l	2065(1239-2955)	2170(1361-3396)	0.253
Peak CK-MB, U/l	189(122-286)	225(134-360)	0.149
Ejection Fraction,%	47±10	50±9	0.036
Wall Motion Score Index	1.59±0.36	1.57±0.34	0.584
IRA, n(%)
LAD	65(58.0)	113(56.2)	0.782
CX	8(7.1)	19(9.5)
RCA	39(34.8)	69(34.3)
Gensini	39(24-55)	38(18-52)	0.213
EF <40%, n(%)	21(18.8)	31(15.4)	0.448
Death, n(%)	38(33.9)	70(34.8)	0.873

CK: creatine kinase; CX: circumflex artery; EF: ejection fraction; IRA: infarct-related artery; LAD: left anterior descending artery; MI: myocardial infarction; RCA: right coronary artery. Continuous variables were presented as mean±SD or median (IQR 25-75); categorical variables were presented as number (%).

During follow-up, death occurred in 108 (34.5%) patients; 38 (33.9%) deaths occurred among non-smokers; 70 (34.8) deaths occurred among smokers (p=0.873). Table 3 shows the study variables in alive and dead patients. In the dead group, patients were older, had lower hemoglobin levels at admission, lower EF and higher WMSI after the ischemic event. Being female, a heavy smoker and presenting EF<40% were also more prevalent factors in the dead group. Table 4 demonstrates the multivariate Cox regression analysis for all-cause mortality. Older age, smoking and having an EF<40% after the infarction were the independent predictors of very long-term mortality. The relative risk for mortality was increasing with a dose-dependent manner in smokers compared to non-smokers. (Group 2 vs. Group 1; HR:1.141; 95%CI: 0.599 to 2.171, Group 3 vs. Group 1; HR:2.130; 95%CI: 1.236 to 3.670, Group 4 vs. Group 1; HR: 2.602; 95%CI: 1.461 to 4.634). Survival curves of smokers and non-smokers are shown in Figure 1 . In Figure 1A, survival curves of smokers and over all non-smokers were presented. In Figure 1B, the unadjusted analysis revealed that Group 2 was exhibiting a better survival curve than Group 1. However, after adjusting for age (Figure 1C) and all other confounders (age, hypertension, diabetes, hemoglobin, infarct-related artery, infarction site, receiving urgent reperfusion treatment and presence of a depressed left ventricle EF) (Figure 1D), the survival curves exhibited the dose-dependent increased risk of smoking.

Table 3

Variable (number)	Alive (205)	Dead (108)	p
Age, years	54±10	62±10	<0.001
Gender (Female)	23(11.2)	23(21.3)	0.017
Hypertension	55(26.8)	43(39.8)	0.019
Diabetes	28(13.7)	18(16.7)	0.475
Family History for CAD	45(22.0)	22(20.4)	0.746
Tobacco Exposure Grade			0.021
Non smoker	74(36.1)	38(35.2)
<20 pack-years	50(24.4)	16(14.8)
20-40 pack-years	62(30.2)	32(29.6)
>40 pack-years	19(9.3)	22(20.4)
Hemoglobin, g/dl	14.6±1.5	14.2±1.8	0.027
WBC*103	11.4±3.6	11.7±4.2	0.471
Creatinine, mg/dl	1.04±0.21	1.07±0.27	0.216
Total cholesterol, mg/dl	194±44	185±40	0.096
LDL, mg/dl	126±37	121±36	0.336
HDL, mg/dl	41±10	41±13	0.683
Triglyceride, mg/dl	115(82-175)	106(73-146)	0.055
Admission Blood Glucose, mg/dl	132±63	137±49	0.566
Anterior MI	118(57.6)	59(54.6)	0.619
Thrombolysis + Primary PCI	163(79.5)	81(75)	0.360
Peak CK, U/l	2156(1308-2999)	2172(1368-3726)	0.269
Peak CK-MB, U/l	202(118-300)	232(141-379)	0.050
Ejection Fraction, %	50±9	47±10	0.016
Wall Motion Score Index	1.54±0.31	1.64±0.40	0.013
IRA			0.979
LAD	117(57.1)	61(56.5)
CX	18(8.8)	9(8.3)
RCA	70(34.1)	38(35.2)
Gensini Score	38(19-52)	38(21-57)	0.396
EF <40%	25(12.2)	27(25.0)	0.004

CAD: coronary artery disease; CK: creatine kinase; CX: circumflex artery; EF: ejection fraction; HDL: high density lipoprotein; IRA: infarct related artery; LAD: left anterior descending artery; LDL: low density lipoprotein; MI: myocardial infarction; RCA: right coronary artery; WBC: white blood cell count. Continuous variables were presented as mean±SD or median (IQR 25-75); categorical variables were presented as number (%).

Table 4

	Exp(B)	95%CI for Exp(B)		p
		Lower	Upper
Age	1,063	1,040	1,088	0,000
Gender (Female)	1,730	0,985	3,040	0,056
Diabetes Mellitus	1,264	0,740	2,158	0,391
Hypertension	1,184	0,765	1,832	0,448
Smoking				0,003
Group 2 vs. Group 1	1,141	,599	2,171	0,689
Group 3 vs. Group 1	2,130	1,236	3,670	0,006
Group 4 vs. Group 1	2,602	1,461	4,634	0,001
Hemoglobin	0,978	0,856	1,118	0,749
Infarction Wall (non-anterior vs anterior )	0,771	0,257	2,307	0,641
Urgent reperfusion (acute presentation vs subacute presentation)	0,978	0,622	1,538	0,924
IRA				0,825
RCA vs LAD	1,305	0,428	3,982	0,640
CX vs LAD	1,089	0,313	3,790	0,894
EF<40%	1,967	1,216	3,181	0,006

CX: circumflex artery; EF: ejection fraction; IRA: infarct-related artery; LAD: left anterior descending artery; RCA: right coronary artery. Group 1: never smokers, Group 2: <20 package*year smokers, Group 3: 20-40 package*year smokers, Group 4: >40 package*year smokers.

Figure 1

The main finding of this study is that among patients who survived after STEMI, those who smoked before the infarction presented higher all-cause mortality than those who had never smoked. The relative risk for mortality grows with the increasing cigarette exposure. The other independent predictors are age and presence of a post-MI depressed left ventricle systolic function.

Many epidemiological studies have demonstrated that the risk associated with smoking increases in a dose-dependent manner.^{10 , 11} Nonetheless, it has been suggested that smokers have a better prognosis in some clinical situations. In addition to acute MI patients, the presence of a smoker’s paradox has been defined in patients with ischemic stroke, resuscitated in-hospital cardiac arrest patients, and TAVI patients.^{12 - 14} The presence of smoker’s paradox in patients with acute MI is controversial. The first studies about the smoker’s paradox were published in the thrombolytic era. Some of those studies suggested that smoking was independently associated with favorable outcomes, and others revealed that smokers had favorable outcomes only in univariate analysis.^{15 , 16} The structure of the culprit lesion in smokers (a larger thrombus burden in smokers than in non-smokers) had been discussed as a potential explanation for the smoker’s paradox in acute MI patients. It was suggested that culprit thrombus in smokers have a tendency towards spontaneous lysis and/or response in thrombolytic therapy, which would be better in smokers compared to non-smokers.¹⁷

The results of the studies during primary PCI were also controversial. Some studies showed that smokers and non-smokers had similar mortality rates; some other studies suggested that smoking was associated with favorable outcomes.^{18 , 19} Also, there were studies showing that smokers present worse outcomes.²⁰ For example, it was demonstrated that smoking was an independent predictor of lower in-hospital mortality in patients with acute MI, even after multiple analysis to control for potential confounders.²¹ Another study found that smokers present lower acute inflammatory response, better microvascular reperfusion and better 30-day mortality rates in the setting of acute MI.³ Clopidogrel responsiveness in smokers has been the most popular mechanism proposed for smoker’s paradox in the primary PCI era. Lower platelet reactivity was observed in smokers treated with clopidogrel in comparison to non-smokers treated with clopidogrel.^{22 , 23} It was suggested that the regional disparities for smoker’s paradox presented and this hypothesis were based on the possible genetic variability of the metabolism of clopidogrel in different races.⁸

Although there are many studies approaching the smoker’s paradox, there are plenty of well-designed analyses which demonstrated opposite findings. Five-year results of the SYNTAX trial demonstrated that smoking was associated with poor outcomes after revascularization in patients with complex coronary artery disease.²⁴ The ACUITY trial established that among patients who were diagnosed with non-ST elevation MI, those who smoked presented a higher one-year mortality rate than those who did not smoke.²⁵ The mortality rate in smokers was found to be significantly higher than in non-smokers in a large registry of STEMI patients who were treated with primary PCI.²⁶ Additionally, cardiac MRI studies revealed that among STEMI patients, those who smoked had higher rates of myocardial hemorrhage and worse cardiovascular outcomes.²⁷ Recent studies rejected the presence of smoker’s paradox in patients with heart failure and in patients with acute ischemic stroke.^{28 , 29} And finally, the result of the study published by Kim et al.³⁰ was quite remarkable. They found that the difference of clopidogrel responsiveness in smokers was largely attributable to the difference in hemoglobin levels. They found similar platelet reactivity between smokers and non-smokers after adjusting the influence of hemoglobin on platelet reactivity.³⁰

In our study, the unadjusted analysis revealed that mild smokers had a better survival curve than non-smokers. This can lead to the idea that mild smoking may be a good habit, but after adjustments according to confounding factors, smoking increased the risk of mortality in a dose-dependent manner. So, baseline confounding factors are very important for the smoker’s paradox. As we mentioned before, many of the studies revealed that the favorable outcomes attributed to smoking vanished after multivariate analysis. The same situation has also been proven for clopidogrel responsiveness.³⁰

Another aspect of smoker’s paradox studies regards design and methodology. When we look at the design and the statistical methods of the studies claiming there is a smoker’s paradox, they provide in-hospital or relatively short-term follow-up results, and the vast majority used the logistic regression analysis, which does not consider the effect of time intervals to the event and cannot handle time-dependent covariates. We could find only one study which used the Cox regression analysis and found favorable outcomes for smoking;³¹ on the other hand, most studies that found unfavorable outcomes for smoking used the Cox regression analysis. And finally, none of the studies were designed to evaluate the smoker’s paradox primarily. This design has always the risk of undetected potential confounders.

The most important limitation of this study is the retrospective design. Adherence to the medical treatment was not known after discharge. We did not know which patients continued to smoke after the STEMI, so we could not draw a conclusion on the effect of ongoing smoking on mortality. We did not have the information about cardiovascular endpoints, like recurrent MI or hospitalization for heart failure.

Smoking gradually increases the risk of all-cause mortality with after STEMI.