Comparison of the survival and neurological outcomes in OHCA based on smoking status: investigation of the existence of the smoker’s paradox

1. Introduction

Smoking is a known risk factor for cardiovascular disease. However, smokers have a rather low myocardial infarction- and stroke-related mortality rate, and this phenomenon, the so-called smoker’s paradox, has also been reported in the case of cardiac arrest. Pollack et al. [1] reported that smokers had good neurological outcomes after out-of-hospital cardiac arrest (OHCA). This effect was significant even after adjusting for confounders. Lahmann et al. [2] also reported that former smoking was associated with improved survival in patients with OHCA. Further, Gupta et al. [3] reported that smokers had a higher rate of survival to discharge and more favorable neurological outcomes following in-hospital cardiac arrest. The mechanism by which smoking improves outcomes after cardiac arrest is not clear but can potentially be explained by the phenomenon of ischemic preconditioning [4].

Recent studies have reported contradictory results regarding the smoker’s paradox in several disease entities. According to a systematic review of the smoker’s paradox in acute coronary syndrome, the findings of only six studies were consistent with the paradox among the 17 included studies [5]. This paradox has not been demonstrated in more recent studies involving routine early invasive therapeutic strategies. Another systematic review of 18 studies on patients with ischemic stroke indicated that smoking is not a protective factor and that the smoker’s paradox is not an actual phenomenon [6]. A recent study of all the incident OHCAs in San Francisco over a 4-year period found no difference in tobacco use rates between survivors and non-survivors [7]. Therefore, the smoker’s paradox should also be evaluated in patients with cardiac arrest through additional research.

The purpose of this study was to evaluate the association between an individual’s smoking status and OHCA-related survival and neurological outcomes by analyzing data from a retrospective multicenter registry.

2. Materials and methods

2.5 Statistical analysis

Three comparisons were performed (Fig. 1). First, the study variables and outcomes were compared among the CS, FS, and NS groups (comparison 1). Second, we compared the CS and non-smoker groups (comparison 2). A non-smoker was defined as a combination of an FS and NS. Third, we compared ever-smokers and NS (comparison 3). An ever-smoker was defined as a combination of a CS and an FS. Descriptive statistics are presented as mean and standard deviation (SD) for continuous variables and as frequency (percentage) for categorical variables. An independent sample t-test or analysis of variance (ANOVA) was performed for continuous variables, and a chi-square test or Fisher’s exact test, for categorical variables. Bonferroni correction was used for post-hoc comparison, and it was judged to have statistical significance when the p-value was 0.0167 or less.

Fig. 1.Patient flow chart. KoCARC, Korean Cardiac Arrest Research Consortium; PSM, propensity score matching.

To adjust for baseline differences between the CS and non-smoker groups and between the ever-smoker and NS groups, propensity scores (PS) were estimated using a logistic regression model as a function of all the potential confounders listed in Table 1. The PS of an individual was defined as the probability of being in the exposure group, given all relevant covariates. Based on the PS, the patients in the smoking status-related groups were matched using the greedy matching algorithm at a 1:1 ratio to create a PS-matched population. The balance in baseline characteristics among the PS-matched cohorts was assessed using a paired t-test or McNemar’s test. Three additional PS-based methods were performed: (i) stratification to divide the sample into five strata based on rank-ordered PS, followed by comparisons between groups within each stratum; (ii) stabilized inverse probability treatment weighting (IPTW) to weight cases by the inverse of the PS; and (iii) regression adjustment to include PS as an additional covariate in a regression model [11, 12].

Table 1.Baseline characteristics among current smoker, former smoker, and never smoker.

Characteristics	Overall	Current smoker	Former smoker	Never smoker	p-value	Post hoc analysis
	(n = 4443)	(n = 885)	(n = 676)	(n = 2882)		CS vs. FS	CS vs. NS	FS vs. NS
Age, years	64.0 $\pm$ 20.4	58.0 $\pm$ 13.6	70.6 $\pm$ 12.6	64.3 $\pm$ 22.9	$<$0.001***	$<$0.001***	$<$0.001***	$<$0.001***
Male sex	2867 (64.5)	815 (92.1)	626 (92.6)	1426 (49.5)	$<$0.001***	0.706	$<$0.001***	$<$0.001***
Hypertension	1974 (45.3)	348 (40.9)	361 (54.1)	1265 (44.6)	$<$0.001***	$<$0.001***	0.056	$<$0.001***
Diabetes	1207 (27.8)	209 (24.7)	227 (34.1)	771 (27.3)	$<$0.001***	$<$0.001***	0.131	$<$0.001***
Dyslipidemia	258 (6.1)	58 (7.0)	55 (8.6)	145 (5.2)	0.002**	0.257	0.048	0.001***
Witnessed arrest	2830 (64.3)	586 (66.7)	451 (67.2)	1793 (62.9)	0.025*	0.845	0.036	0.035
Bystander CPR	2266 (51.0)	457 (51.6)	343 (50.7)	1461 (50.7)	0.701
Shockable rhythm	953 (21.4)	315 (35.6)	153 (22.6)	485 (16.8)	$<$0.001***	$<$0.001***	$<$0.001***	0.018
Bystander AED use	75 (1.7)	12 (1.4)	19 (2.9)	44 (1.6)	0.048*	0.046	0.752	0.022
Prehospital defibrillation	1202 (27.6)	377 (43.4)	185 (27.7)	640 (22.7)	$<$0.001***	$<$0.001***	$<$0.001***	0.006**
Prehospital epinephrine	533 (12.0)	111 (12.6)	96 (14.2)	326 (11.4)	0.100
Prehospital ROSC	801 (18.4)	266 (30.6)	114 (17.1)	421 (14.9)	$<$0.001***	$<$0.001***	$<$0.001***	0.166
Coronary angiography	780 (17.6)	279 (31.5)	129 (19.1)	372 (12.9)	$<$0.001***	$<$0.001***	$<$0.001***	$<$0.001***
TTM	592 (15.4)	175 (22.6)	85 (14.4)	332 (13.4)	$<$0.001***	$<$0.001***	$<$0.001***	0.530
Survival to discharge	841 (18.9)	269 (30.4)	126 (18.6)	446 (15.5)	$<$0.001***	$<$0.001***	$<$0.001***	0.043
CPC 1 or 2	601 (13.5)	212 (23.9)	87 (12.9)	302 (10.5)	$<$0.001***	$<$0.001***	$<$0.001***	0.073
CS, current smoker; NS, never smoker; FS, former smoker; CPR, cardiopulmonary resuscitation; AED, automated external defibrillator; ROSC, return of spontaneous circulation; TTM, targeted temperature management; CPC, cerebral performance category. * $<$0.05, ** $<$0.01, *** $<$0.001.

The association between outcomes and smoking status was assessed using a logistic regression analysis. We first performed univariable logistic regressions on outcome predictors of OHCA, such as age; sex; history of hypertension, diabetes mellitus, or dyslipidemia; initial ECG rhythm; witnessed arrest; bystander CPR; bystander AED use; coronary angiography findings; and TTM. We then built a stepwise multivariable regression model starting with the predictor that had the strongest association with the outcome based on the results of the univariable logistic regression. Collinearity between the predictors was assessed by calculating the variance inflation factor (VIF) for each predictor and excluding variables with a VIF greater than 5. No predictor met this criterion, and therefore none were removed from the analysis based on collinearity. Adjusted odds ratios (AOR) and 95% confidence intervals (CI), obtained from the multivariable logistic regression after controlling for potential confounders, were used to interpret associations between variables and outcomes.

Data analyses and visualization were performed using Excel (Microsoft, Redmond, WA, USA) and the SAS program (version 9.4, SAS Institute Inc., Cary, NC, USA). A p -value of $<$0.05 was considered statistically significant.

3. Results

Of the 9521 OHCA patients listed in the KoCARC registry from October 2015 to June 2019, we enrolled 4443 patients in the final analysis after excluding patients whose smoking history could not be obtained. The overall survival rate to discharge was 18.9%, and the proportion of patients with good neurological outcomes was 13.5%. Among these patients, 885 (19.9%) were CS, 676 (15.2%) were FS, and 2882 (64.9%) were NS (Fig. 1).

The baseline characteristics of the overall population and the three groups according to the smoking status are presented in Table 1. CS had a significantly better survival to discharge rate (30.4% vs. 18.6% vs. 15.5%, p $<$ 0.001) and more favorable neurological outcomes (23.9% vs. 12.9% vs. 10.5%, p $<$ 0.001) than FS or NS. However, the baseline characteristics, apart from bystander CPR and prehospital epinephrine use, were also significantly different among the groups. After the PS matching procedure, 639 matched pairs were selected between the CS and non-smokers, and 980 matched pairs were selected between the ever-smokers and NS. The differences in baseline characteristics were well balanced between the PS-matched cohorts, but significant differences in resuscitation outcomes were not observed after PS-matching between CS and non-smokers (survival to discharge: 31.8% vs. 28.5%, p = 0.136; favorable neurological outcomes: 25.4% vs. 22.7%, p = 0.172, Table 2) and between ever-smokers and NS (survival to discharge: 23.4% vs. 21.5%, p = 0.241; favorable neurological outcomes: 17.1% vs. 15.7%, p = 0.294, Table 3). The survival to discharge rate among the groups with various smoking statuses is presented in Fig. 2.

Fig. 2.Rate of survival to hospital discharge among patients with different smoking status according to the comparisons. PSM, propensity score matching.

In the unadjusted logistic regression model, CS had significantly higher chances of survival to discharge (OR, 2.28; 95% CI, 1.925–2.699; p $<$ 0.001) and better neurological outcomes (OR, 2.566; 95% CI, 2.129–3.094; p $<$ 0.001) than non-smokers. Ever-smokers also had significantly higher chances of survival to discharge (OR, 1.85; 95% CI, 1.589–2.155; p $<$ 0.001) and better neurological outcomes (OR, 2.024; 95% CI, 1.702–2.408; p $<$ 0.001) than NS. However, significant improvements in outcomes were not observed in the multivariable adjusted models in both CS vs. non-smokers and ever-smokers vs. NS (Table 4).

A sensitivity analysis was performed using different statistical methods with PS. There were no significant differences in outcomes between ever-smokers and NS on constructing PS-adjusted models by stratification, regression adjustment, and IPTW. When comparing CS and NS, regression adjustment with PS and IPTW did not reveal statistically significant findings; however, the neurological outcomes of CS were more favorable than those of non-smokers when using the stratification method (Table 4).

4. Discussion

The results of this study suggest that the smoker’s paradox may be a pseudo-paradox in OHCA patients. Although current smoking and ever-smoking were associated with improved survival and favorable neurological outcomes following OHCA, the protective effects of smoking often disappeared after various statistical adjustments, including multivariable logistic regression or PS-based adjustments.

A previous study that investigated the smoker’s paradox in OHCA revealed better outcomes in smokers than in non-smokers, even after adjusting for confounders [11]. However, the sample size of the study was relatively small, comprising of 95 smokers and 86 non-smokers. A study of patients with in-hospital cardiac arrest, which analyzed a large nationwide dataset, showed that smokers had a higher rate of survival to hospital discharge and more favorable neurological outcomes than non-smokers [3]. There is a possibility of systematic error owing to the finding that smokers who experience an acute cardiac event could have a greater case fatality before hospital admission than non-smokers [13, 14]. Those admitted alive to the hospital would already represent the survivors.

These two studies, which investigated the smoker’s paradox in cardiac arrest, only compared ever-smokers with NS, but we added a comparison between CS and current non-smokers, including FS. If smoking exerts a protective effect in OHCA patients through the mechanism of ischemic preconditioning, a dose-response relationship may be possible among CS, FS, and NS. However, the FS group had the oldest patients and the largest proportion of patients with histories of hypertension or diabetes. CS had a higher survival rate and more favorable neurological outcomes than non-smokers, and both these outcomes were the same between ever-smokers and NS. However, these differences were not statistically significant after PS matching.

This was also the case in the logistic regression analysis. As a result of multivariable logistic regression, the statistical differences observed in the unadjusted model for survival or neurological outcomes in both CS and ever-smokers compared with in non-smokers or NS were no longer observed. We performed additional PS-adjusted analyses, such as independent 1:1 PS matching, IPTW, and regression adjustment, all of which showed the same results. However, in the case of the stratification method, it was found that the neurological outcomes were significantly better in CS than in non-smokers (OR, 1.348; 95% CI, 1.050–1.731; p = 0.019). This implies that the results may vary depending on the statistical adjustment method used.

The smoker’s paradox was first described in 1968 by Weinblatt et al. [15] who stated that smokers had lower mortality after myocardial infarction compared to non-smokers. Some studies have supported the existence of this phenomenon, whereas others have refuted its existence. This phenomenon, which has more unfavorable outcomes after cardiac arrest in non-smokers than in smokers, explains that frequent hypoxic injury due to smoking can diminish the effect of reperfusion injury during post-cardiac arrest period. The concept of “ischemic preconditioning of smoking” is that mortality rate of myocardial infarction is lower in smokers, although smoking increases the prevalence of acute coronary syndromes [16, 17]. The apparent smoker’s paradox has usually been disproved in many studies on acute coronary syndrome after the adjustment for measured risk factors [5]. For example, among myocardial infarction patients undergoing percutaneous coronary interventions, FS and CS had a decreased unadjusted hazard ratio for both 30-day and 1-year mortality when compared with NS. However, this protective effect of smoking was not observed after adjusting for potential confounding factors [18]. Similar controversial findings exist in the case of patients with ischemic stroke. Smokers may exhibit better recovery and thrombolytic responses than non-smokers.

The smoker’s paradox has also been reported in patients with conditions other than myocardial infarction, such as stroke [19], trauma [20] and heart failure [21]. However, this paradox has also been denied in recent studies on such patients. A systematic review of ischemic stroke reported that smoking was not a protective factor and that the smoker’s paradox was not an actual phenomenon [6]. Among adolescent trauma patients, smokers had an increased rate of pneumonia, a longer length of hospital stay, and no difference in overall mortality compared to non-smokers [22]. There was no survival advantage of smoking in patients hospitalized with heart failure, and the smoker’s paradox is likely to be a result of residual confounding [23].

In this study, statistical adjustment was attempted using various methods, and the smoker’s paradox was not observed with most of these methods except for with the stratification method using PS. The neurological outcomes were more favorable in CS than in non-smokers when statistical significance was maintained even after the adjustment by stratification with PS. We divided the data into five strata, and the neurological outcomes in CS were better in the first and last quintiles. Therefore, the results indicating whether the smoker’s paradox occurred are likely to differ depending on the statistical adjustment method, and it is necessary to conduct a thorough adjustment for various confounders.

This study has several limitations. First, as a retrospective study using a registry, there are inherent limitations, including reporting bias, coding errors, and missing data. More than half of the patients were excluded due to the lack of information regarding their smoking history. Second, smoking status was usually reported by family members who may have incorrectly recorded individuals as non-smokers. Ideally, an objective test, such as urinary nicotine or cotinine tests, could have been used; however, this is impractical in the OHCA setting. Third, the duration and amount of smoking was not reported and dose-dependent effects of smoking were not considered. Even among smokers, the prognosis according to smoking period might be different. Lastly, we recommend applying caution in extrapolating these findings to Caucasian and/or other populations because this study was based on the Korean population.

5. Conclusions

This study suggests that smoking is not an independent prognostic factor in OHCA. The statistically significant better outcomes observed in CS or ever-smokers were not maintained after adjusting for confounders. Therefore, the smoker’s paradox should be investigated through additional prospective studies on OHCA and should not be interpreted as a benefit of or justification for smoking.

Abbreviations

OHCA, out-of-hospital cardiac arrest; CS, current smoker; FS, former smoker; NS, never-smoker; KoCARC, Korean Cardiac Arrest Research Consortium; CPR, cardiopulmonary resuscitation; ECG, electrocardiography; AED, automated external defibrillator; ROSC, return of spontaneous circulation; TTM, targeted temperature management; CPC, cerebral performance category; IPTW, inverse probability treatment weighting; PS, propensity score; OR, odds ratio; CI, confidence interval.

Author contributions

Conceptualization—SPC, GY. Data curation—HSL, GY. Formal analysis—SPC, HSL. Investigation—SPC, GY. Methodology—TK, JSY. Software—HSL, GY. Validation—YSP. Visualization—SPC. Writing — original draft—SPC, GY. Writing — review & editing—SPC, TK, HSL, YSP, JSY.

Ethics approval and consent to participate

The study protocol was reviewed and approved by the institutional review boards (IRB) of the 62 participating hospitals in the KoCARC OHCA registry. The need for informed consent was waived by the IRB. This research project was registered at ClinicalTrials.gov (identifier: NCT03222999). Patient consent was waived due to observational nature of the study.

Acknowledgment

We would like to acknowledge and thank the investigators from all participating hospitals of KoCARC: Woon Yong Kwon (Seoul National University Hospital), Sang Kuk Han, Phil Cho Choi (Kangbuk Samsung Medical Center), Sang O Park, Jong Won Kim (Konkuk University Medical Center), Han Sung Choi, Jong Seok Lee (Kyung Hee University Hospital), Sung Hyuk Choi, Young Hoon Yoon (Korea University Guro Hospital), Su Jin Kim (Korea University Anam Hospital), Min Seob Sim, Gun Tak Lee (Samsung Medical Center), Shin Ahn (Asan Medical Center), Jong Whan Shin (SMG-SNU Boramae Medical Center), Sang Hyun Park, Keun Hong Park (Seoul Medical Center), In Cheol Park, Tae Young Kong (Yonsei University Gangnam Severance Hospital), KyoungWon Lee, Chu Hyun Kim (Inje University Seoul Paik Hospital), Youngsuk Cho (Hallym University Kangdong Sacred Heart Hospital), Gu Hyun Kang, Yong Soo Jang (Hallym University Kangnam Sacred Heart Hospital), Tai Ho Im, Jae Hoon Oh (Hanyang University Seoul Hospital), Seok Ran Yeom, Sang Kyoon Han (Pusan National University Hospital), Jae Hoon Lee (Dong-A University Hospital), Jeong Bae Park, Hyun Wook Ryoo (Kyungpook National University Hospital), Kyung Woo Lee, Tae Chang Jang (Daegu Catholic University Medical Center), Jae-hyug Woo (Gachon University Gil Medical Center), Woon Jeong Lee, Seon Hee Woo (The Catholic University of Korea Incheon St. Mary’s Hospital), Sung Hyun Yun, Tae Jin Cho (Catholic Kwandong University International St. Mary’s Hospital), Sun Pyo Kim, Yong Jin Park (Chosun University Hospital), JinWoong Lee, Wonjoon Jeong (Chungnam National University Hospital), Sung Soo Park, Jae Kwang Lee (Konyang University Hospital), Ryeok Ahn, Wook Jin Choi (Ulsan University Hospital), Young Gi Min, Eun Jung Park (Ajou University Hospital), You Hwan Jo, Joong Hee Kim (Seoul National University Bundang Hospital), In Byung Kim (Myongji Hospital), Han Jin Cho (Korea University Ansan Hospital), Seung Cheol Lee, Sang Hun Lee (Dongguk University Ilsan Hospital), Young Sik Kim, Young Rock Ha (Bundang Jesaeng Hospital), Jin Sik Park, MyoungWoo Lee (Sejong Hospital), Dai HanWi (Wonkwang University Sanbon Hospital), Sang Ook Ha,Won Seok Yang (Hallym University Pyeongchon Sacred Heart Hospital), Ok Jun Kim, Tae Nyoung Chung (Cha University Bundang Medical Center), Soon Joo Wang (Hallym University Dongtan Sacred Heart Hospital), Jun Hwi Cho, Chan Woo Park (Kangwon National University Hospital), An Mu Eob, Tae Hun Lee (Hallym University Chuncheon Sacred Heart Hospital), Sang Chul Kim, Hoon Kim (Chungbuk National University Hospital), Han Joo Choi, Chan Young Koh (Dankook University Hospital), JungWon Lee, DongWook Lee (Soonchunhyang University Cheonan Hospital), Tae Oh Jung, Jae Chol Yoon (Chonbuk National University Hospital), Dai Hai Choi, Jung Tae Choi (Dongguk University Gyeongju Hospital), Jin Hee Jeong, Soo Hoon Lee (Gyeongsang National University Hospital), Ji Ho Ryu, Maeng Real Park (Pusan National University Yangsan Hospital), Won Kim (Cheju Halla General Hospital), Sung Wook Song, Woo Jung Kim (Jeju National University Hospital), Joon-myoung Kwon, Eui Hyuk Kang (Mediplex Sejong Hospital), Sang Chan Jin, Tae-kwon Kim (Keimyung University Dongsan Medical Center), Hyuk Joong Choi (Hanyang University Guri Hospital), Seong Chun Kim (Gyeongsang National University Changwon Hospital).

To steering committee, comprised of following individuals: Sung Oh Hwang (Chair,Wonju Severance Christian Hospital), Sang Do Shin (Chair of Steering Committee, Seoul National University hospital), Hyuk Jun Yang (Advisory Committee, Gachon University Gil hospital), Sung Woo Lee (Security and Monitoring Board, Korea University Anam hospital), Kyung Jun Song (Secretariat, SMG-SNU Boramae Medical Center), Seung Sik Hwang (Epidemiology and Prevention Research Committee, Seoul National University), Gyu Chong Cho (Community Resuscitation Research Committee, Hallym University Kangdong Sacred Heart Hospital), Sung Woo Moon (Emergency Medical Service Resuscitation Research Committee, Korea University Ansan Hospital), Kyoung Chul Cha (Hospital Resuscitation Research Committee, Wonju Severance Christian Hospital), Won Young Kim (Hypothermia and Post-Resuscitation Care Research Committee, Asan Medical Center), Sang Hoon Na (Cardiac Care Resuscitation Research Committee, Seoul National University Hospital), Young Ho Kwack (Pediatric Resuscitation Research Committee, Seoul National University hospital).

To member of Secretariat: Joo Yeong Kim (Korea University Ansan hospital), Jeong Ho Park (Seoul National University hospital), Sun Young Lee (Seoul National University hospital), and Jungeun Kim (Seoul National University hospital).

We would also like to thank the National Fire Agency for providing prehospital EMS data, and the Korean Association of Cardiopulmonary Resuscitation (KACPR) for their support.

Funding

This research received no external funding.

Conflict of interest

The authors declare no conflict of interest.