Resolving the Smoking Paradox: No Evidence for Smoking-Induced Preconditioning in Large Vessel Occlusion Stroke

Tobacco smoking is one of the leading causes of stroke and preventable death globally, responsible for many years of life lost [1]. Despite its undisputed role as a risk factor for cardiovascular disease, early studies on myocardial infarction reported better outcomes in patients with a history of smoking [2, 3]. This finding was corroborated in studies on thrombolysis in myocardial infarction [4, 5], and hence the term “smoking paradox” came into being. The smoking paradox was later discovered for stroke, when it was observed that smokers had better recanalization rates [6] and improved early outcomes in acute ischemic stroke when treated with intravenous thrombolysis [7‒10].

One suggested mechanism for better outcome was a preconditioning-like protection (i.e., smoking-induced ischemia tolerance [11]) due to better collaterals and maintained perfusion (see Fig. 1). Alternatively, smoking might be associated with a particular subtype of stroke or indicate stroke patients with a more favorable outcome, such as younger patients [6, 12]. However, previous analyses included heterogenous patient cohorts with different types of vascular occlusion, so bias could have arisen due to different clot locations and varying degrees of stroke severity.

In our study, we aimed to resolve the smoking paradox by overcoming such bias and comparing only patients with middle cerebral artery M1 segment (MCA-M1) occlusions. In order to prevent potential treatment bias, we selected only patients who were submitted to treatment with mechanical thrombectomy (MT). We analyzed differences in stroke severity as well as patient clinical and imaging characteristics according to smoking status. To prove the concept of a smoking-induced vascular and tissue protection during stroke, we analyzed collateral status and brain perfusion on pre-MT imaging and further assessed recanalization rates on post-MT digital subtraction angiography (DSA) images. We aimed to assess whether 1) smoking was associated with less severe strokes and whether 2) collaterals were indeed more extensive or tissue perfusion was better maintained in smokers due to chronic tobacco exposure (preconditioning hypothesis).

The study was performed according to the ethical guidelines of the Canton of Zurich with approval from the Local Ethics Committee (“PREDICT” project; KEK-ZH-Nr. 2014-0304). Written consent was obtained from all subjects or legally authorized representatives as required. The protocol allowed use of data of deceased patients if no objection against use of data for research was documented (according to Article 34 of the Swiss Human Research Law [implemented in January 2014]).

We retrospectively collected data of patients with acute ischemic stroke due to MCA-M1 segment occlusion sent for MT at the University Hospital Zurich between 2014 and 2018. Exclusions were due to omitted consent to use data for research, missing NIHSS score on admission, or lack of information about smoking status. Details on inclusion/exclusion criteria are provided in Figure 2.

Thrombectomy decisions were made by treating stroke physicians according to current clinical guidelines. Standard demographic and clinical data were recorded (see Tables 1, 2). Patients were categorized according to smoking status (current, former, or never) using information available in medical records. Number of pack years and duration of abstinence could not be considered in our analysis.

Stroke imaging (computed tomography angiography (CTA), DSA, and CT perfusion [CTP]) was performed using routine stroke protocols – performed at the University Hospital Zurich or local referring hospitals. To avoid introducing inhomogeneity into core and mismatch imaging analysis (processed with Olea Sphere’s Acute Stroke Care Plug-In [Olea Sphere Version 3.0.22, La Ciotat, France]), we excluded patients with initial magnetic resonance (MR) perfusion and analyzed CTP imaging only. Quantitative (cerebral blood flow [CBF], cerebral blood volume, and mean transit time) and qualitative/semi-quantitative (time-to-peak) maps were calculated from perfusion images. Automatic volume segmentation was used to define the core (rCBF <30%) and penumbral volumes (Tmax threshold at values > 6 s). An experienced neurologist and a neuroradiologist categorized CT or MR angiography data pre-MT according to the EXTEND-IA criteria [13] as good, moderate, and poor; and DSA data according to the modified treatment in cerebral infarction (mTICI) scale [14], including subcategory mTICI 2c [15].

The primary clinical endpoint was the association of smoking with stroke severity on admission; i.e., the National Institutes of Health Stroke Scale (NIHSS) score on admission. Secondary endpoints were infarct core volume (CTP imaging) and collateral status (CTA, DSA).

For descriptive statistics, we calculated median and interquartile range (IQR) (see Table 2 and 4) for continuous variables and tested for group differences using a Kruskal-Wallis test. Categorical variables were expressed as frequencies and percentages; and χ² tests were used to test for differences between smoking groups.

The association between smoking and NIHSS, stroke core volume, and collateral status was assessed using linear and ordinal logistic regression models. All regression models were adjusted for age, sex, admission parameters including systolic blood pressure and creatinine, pre-stroke medication with antihypertensives and antiplatelets, risk factors such as diabetes and peripheral artery disease, previous stroke, and presumed etiology (Trial of ORG 10172 in Acute Stroke Treatment [TOAST] criteria). These factors were selected according to expert knowledge and their potential association to the outcome measures observed in previous studies [16]. Estimates and corresponding 95% confidence intervals (CIs) are reported together with p values. p values <0.05 were considered statistically significant. Statistical analysis was performed in IBM SPSS Statistics (Version 27, 28) and R (Version 4.0.2).

To estimate the required sample size, we used the package pwr in R. We powered for the primary goal to show at least one difference in means across the three smoking groups for the normally distributed variable NIHSS by using an ANOVA. We assumed a power of 80%, a significance level of 0.05 and a medium effect size (0.25). This yielded a required sample size of 159 (53 per group).

We screened 2,980 patients with acute ischemic stroke from 2014 to 2018 and identified 401 patients with MCA-M1 occlusion treated with MT. Ultimately, we were able to include 320 patients in our analysis. Exclusions were due to consent to use data for research not being given (n = 65), missing NIHSS score on admission (n = 2), or lack of information about smoking status (n = 14; see Fig. 2).

Median age for all patients was 72.1 years and 50.9% were female (see Table 1 for unadjusted results). Sixty-three patients (19.7%) were classified as current smokers, 60 (18.8%) as former smokers, and 197 (61.6%) as never smokers. Smokers were younger on average (median: 62 years, range: 25–89 years) compared to former smokers (median: 70.9 years, range: 46–90 years) and never smokers (median: 77.3 years, range: 19–97 years; p < 0.001). Current and former smokers were significantly more often male (61.9% and 65%) while never smokers were more often female (59.9%; p < 0.01, see Table 1).

Most common risk factors were hypertension (67.5%), dyslipidemia (50%), and atrial fibrillation (38.8%). Atrial fibrillation occurred more often in never smokers (45.2%) and former smokers (40%) compared to current smokers (17.5%; p < 0.01). Peripheral artery disease was more common in former smokers (10%) and current smokers (6.3%) compared to never smokers (2%; p = 0.019). Regarding history of cerebrovascular events (previous stroke, transient ischemic attack, or intracerebral hemorrhage); no significant differences between smoking groups were found.

Median NIHSS scores on admission did not differ significantly between groups (never smokers 14 [IQR 7], former smokers 15 [IQR 9], current smokers 11 [IQR 8]; p = 0.09, see Table 2 and online suppl. Figure S3; for all online suppl. material, see https://doi.org/10.1159/000533436). In accordance with their higher age at stroke, former smokers more often suffered from pre-stroke disability (modified Rankin Scale [mRS]; p = 0.002). However, disability after 3–6 months was similar in all groups (median mRS 3 for current and never smokers [IQR 3 and 4, respectively]; median mRS 2.5 for former smokers [IQR 5]; p = 0.185).

The most common presumed cause of stroke overall was cardioembolic (TOAST II; 42.8%), followed by undetermined origin (TOAST V; 35.6%) and large vessel disease (TOAST I; 14.7%). Current and former smokers more often showed large vessel disease (TOAST I; 27% and 18.3%) compared to never smokers (9.6%; p = 0.017).

Occurrence of in-hospital complications such as symptomatic stroke or intracerebral hemorrhage did not differ significantly between groups. However, more never (12.7%) and former smokers (13.3%) passed away during their hospitalization than current smokers (1.6%; p = 0.036).

There was no significant difference regarding collateral status between groups (see Table 3). Successful recanalization (mTICI scores 2b, 2c, and 3) was achieved in 85.3% of all patients (p = 0.856), with a similar distribution between groups. However, when looking at individual mTICI scores, never and former smokers more often gained complete reperfusion (mTICI 3; 52% of all never smokers and 46.7% of all former smokers) than current smokers (27%); while current smokers were more often classified with mTICI 2b (60.3%) than former (36.7%) and never smokers (32.7%; p = 0.001). Please note that mTICI 2c was achieved in 1 patient only. CTP imaging analysis showed no significant group difference regarding median volume of stroke core (current smokers 29.5 cm³ [IQR 5.7], never smokers 33.9 cm³ [IQR 4.2], former smokers 43.1 cm³ [IQR 4.3]), median stroke penumbra (current smokers 89.7 cm³ [IQR 5.2], never smokers 90.3 cm³ [IQR 4.0], former smokers 102.6 cm³ [IQR 4.5]), or any of the perfusion parameter maps (see Table 4).

Linear regression analysis adjusted for potential co-variables showed no statistically significant association of smoking with stroke severity on admission (see Table 5). Never and former smoking status were not associated with NIHSS scores on admission (estimate of 0.54, 95% CI: −1.27–2.35, p = 0.557 for never smokers and estimate of 0.56, 95% CI: −1.53–2.64, p = 0.598 for former smokers). Only previous stroke was significantly associated with higher NIHSS on admission (estimate of 2.47, 95% CI: 0.19–4.75, p = 0.034).

The ordinal logistic regression analysis showed no statistically significant association of smoking or any other variable with collateral status (see online suppl. Table S1). Furthermore, the linear regression analysis showed no statistically significant association between smoking status or any other variable with stroke core volume (see online suppl. Table S2).

The results of our study do not support the existence of a “smoking paradox” in stroke, i.e., a better outcome after stroke for patients who smoke due to stroke-induced protection. Neither stroke severity nor collateral status or perfusion parameters were suggestive of a preconditioning effect provided by smoking (current or former smoking status). In this cohort of stroke patients with M1 occlusion, smokers were younger, male, and more often diagnosed with large vessel disease (TOAST I). Both younger age and atherosclerotic stroke cause are associated with a better prognosis after stroke [17, 18]. However, despite their younger age and less common cardiac stroke etiology, smokers did not have a better outcome or more favorable reperfusion success, arguing once more against the preconditioning hypothesis. Our study indicates that, in unadjusted descriptive analysis, smoking is associated with certain patient and stroke characteristics (younger age, male sex, and atherosclerotic stroke); favoring less severe strokes.

These findings are in line with previous work [6, 10, 12, 19] investigating the smoking paradox in stroke. Similar to our results, the association between smoking and good outcome lost significance [6, 20] when adjusting for confounders such as age and sex. While previous studies had indicated that collateral supply did not differ between smokers and nonsmokers [10, 19], we added the analysis of infarct core and mismatch volume as potential correlates of smoking-induced preconditioning (see Fig. 1). However, volume of the infarct core and mismatch were similar in all patient groups. While the results of our study corroborate previous reports doubting the existence of a truly beneficial (preconditioning) effect of smoking on the cerebrovascular system, we aimed to further dissect a potential adaptation of the vascular network in response to tobacco exposure: a more favorable perfusion profile (small core, large mismatch), better collaterals, or more successful recanalization. None of this was found in our selected patient cohort with similar vascular occlusions (MCA-M1 occlusions; all deemed good candidates for MT).

Assignment of patients to different groups according to smoking status was difficult as information about number of cigarettes per day/pack years or years of abstinence was usually not available in medical records. The classification of a patient as a current, former, or never smoker depended on the individual doctor responsible for the patient’s medical management at the time of administration and thus his/her recordings. Investigating a dose-response relationship between smoking and stroke severity on admission was not possible.

Furthermore, our study was a single-center retrospective analysis. Thus, without a randomized controlled study design, we could not derive unequivocal conclusions about causality with respect to smoking. Selecting a group of patients with similar vascular occlusions and treatment limited our sample size, and may reduce generalizability of our results, but it was crucial for overcoming bias regarding stroke type.

The availability of DSA data for patients sent for treatment with MT allowed a thorough assessment of recanalization success. Only 1 patient within our cohort was classified with an mTICI score of 2c. We attributed this to the fact that this subcategory was rather new and not widely used in the years of the data collection. Image quality in the earlier years of our study might also have been insufficient for a confident visual classification of a near-perfect result. Lastly, this single case might have been a coincidence.

We demonstrate no advantage of tobacco smoking that would support the existence of a smoking paradox in stroke. On the contrary; smoking puts patients at risk for stroke at a younger age, thus increasing suffering and disability across the life span. Thus, with respect to the paramount importance of stroke prevention and implementation of preventive strategies including campaigns against smoking, we suggest closing the chapter of the enigmatic “smoking paradox” in stroke.

This study protocol was reviewed and approved by the Ethics Committee of the Canton of Zurich, Switzerland (“PREDICT” project; approval number KEK-ZH-Nr. 2014-0304). Written consent was obtained from all subjects or legally authorized representatives as required. The protocol allowed use of data of deceased patients if no objection against use of data for research was documented (according to Article 34 of the Swiss Human Research Law [implemented in January 2014]).

The authors have no conflicts of interest to declare.

Research activity of S.W., L.H., and M.E.A. was financially supported by the Swiss National Science Foundation (PP00P3_202663 and 310030_200703), the UZH Clinical Research Priority Program Stroke, the Swiss Heart Foundation, and the Baugarten Foundation.

R.E.W. and S.W. conceived and designed the study. R.E.W. wrote the first draft of the manuscript and was responsible for patient recruitment, image analysis, and statistical data analysis. A.B. and J.H. were involved in concept development and image analysis. L.H. performed statistical data analysis. M.E.A., H.S., Z.K., and A.R.L. contributed to data acquisition. S.W. was involved in patient recruitment, image analysis, interpretation of study data, and was responsible for the overall supervision of the project. All authors reviewed and edited the manuscript and approved the final version of the manuscript.