Introduction
More than one in two patients with schizophrenia (SZ) uses tobacco worldwide (de Leon & Diaz, Reference de Leon and Diaz2005). In France, the estimated prevalence is also around 55%, which is almost twice as frequent as in the general French population (Guignard et al., Reference Guignard, Beck, Wilquin, Andler, Nguyen-Thanh, Richard and Arwidson2015; Mallet et al., Reference Mallet, Le Strat, Schürhoff, Mazer, Portalier and Andrianarisoa2017a, Reference Mallet, Le Strat, Schürhoff, Mazer, Portalier and Andrianarisoa2018). There are mainly four hypotheses that are mutually non-exclusive and may collectively contribute to the association of SZ with tobacco smoking.
First, patients with SZ have difficulty with nicotine withdrawal, despite similar motivation to other patients and even in general population (Etter, Mohr, Garin, & Etter, Reference Etter, Mohr, Garin and Etter2004; Lucatch, Lowe, Clark, Kozak, & George, Reference Lucatch, Lowe, Clark, Kozak and George2018). This has been corroborated by a study that compared smokers with SZ to non-psychiatric control smokers in an abstinence condition; the SZ group reported stronger cravings and withdrawal symptoms and had a shorter time to smoking lapse compared to the control group (Tidey, Colby, & Xavier, Reference Tidey, Colby and Xavier2014). These observations may also result from dysexecutive syndrome that would decrease the pursuit of a goal, despite initial motivation (Mallet et al., Reference Mallet, Le Strat, Schürhoff, Mazer, Portalier and Andrianarisoa2018).
One of the other hypotheses frequently mentioned to explain the strong association between SZ and tobacco use is the addiction vulnerability hypothesis. Given that both diseases are highly influenced by genetics, there might be shared genetic components predisposing (propensity) to both smoking behaviours and the development of SZ. Genetic pleiotropy and environmental factors inherent to SZ make these subjects more vulnerable to smoking (Ma et al., Reference Ma, Li, Xu, Wang, Yao, Liu and Li2020). One of the most promising findings is that variants in the CHRNA5/A3/B4 cluster on chromosome 15q24 are associated not only with nicotine dependence but also with SZ. Other genes would be involved according to a recent integrative study (Ma et al., Reference Ma, Li, Xu, Wang, Yao, Liu and Li2020).
A third alternative explanation of the high rate of tobacco smoking in SZ is an attempt of self-medication of extrapyramidal side-effects and cognitive dysfunction. Regarding side-effects, we previously demonstrated that SZ smokers had similar clinical profile (dyskinesia, extrapyramidal symptoms, abnormal involuntary movements) than non-smoker patients (Mallet et al., Reference Mallet, Le Strat, Schürhoff, Mazer, Portalier and Andrianarisoa2017a, Reference Mallet, Le Strat, Schürhoff, Mazer, Portalier and Andrianarisoa2018). On the other hand, several studies have shown an improvement in various cognitive processes following the administration of nicotine, whether in healthy subjects (Heishman, Kleykamp, & Singleton, Reference Heishman, Kleykamp and Singleton2010) or with SZ (Ahlers et al., Reference Ahlers, Hahn, Ta, Goudarzi, Dettling and Neuhaus2014; Hahn et al., Reference Hahn, Harvey, Concheiro-Guisan, Huestis, Holcomb and Gold2013; Sacco et al., Reference Sacco, Termine, Seyal, Dudas, Vessicchio, Krishnan-Sarin and George2005). Cognitive disorders are core features of SZ, impacting daily life and the prognosis of the disease (Kahn & Keefe, Reference Kahn and Keefe2013; Mallet, Le Strat, Dubertret, & Gorwood, Reference Mallet, Le Strat, Dubertret and Gorwood2020). The neurophysiological abnormality consisting of a decrease in the inhibition of the P50 auditory-evoked response to a second pair stimuli has been reported in SZ and their relatives (Freedman et al., Reference Freedman, Coon, Myles-Worsley, Orr-Urtreger, Olincy, Davis and Byerley1997) and is related to sensory gating and attention deficit. This could involve a hippocampal nicotine receptor dysfunction and could be temporally improved by nicotine (Leonard et al., Reference Leonard, Adler, Benhammou, Berger, Breese, Drebing and Freedman2001). Accordingly, a recent systematic review highlighted that a single dose of nicotine could improve a range of cognitive functions in SZ (Dondé et al., Reference Dondé, Brunelin, Mondino, Cellard, Rolland and Haesebaert2020). However, the pro-cognitive effects of nicotine found in several of the studies of this review could not be replicated. Moreover, it is impossible to infer the effects of chronic tobacco use from the results of these experimental short-term studies on cognition performance. Also, it is difficult to differentiate the effects of nicotine and tobacco on cognitive processes. Overall, the hypothesis of self-medication (through the procognitive effect of nicotine) has probably led physicians and psychiatrists to underestimate the major problem of tobacco use in SZ. Still, smokers with SZ are at higher risk for tobacco-related morbidity and mortality. They have about 10–25 years of shortened lifespan, with 53% of this being related to tobacco-smoking conditions (Lucatch et al., Reference Lucatch, Lowe, Clark, Kozak and George2018). Reducing smoking rates could drastically change mortality rates and improve outcomes for these patients. Studies comparing cognition in smokers v. non-smokers in SZ show divergent results and concern only small samples, not allowing the consideration of confounding factors (Lucatch et al., Reference Lucatch, Lowe, Clark, Kozak and George2018). However, a meta-analysis recently provided strong evidence that chronic smoking in SZ was related to cognitive impairment rather than the contrary (Coustals et al., Reference Coustals, Martelli, Brunet-Lecomte, Petillion, Romeo and Benyamina2020). Following this statement, we could assume that if chronic tobacco use was an attempt of enhancing cognitive dysfunction, the corollary hypothesis is that cognitive performances worsen after complete withdrawal. However, a small recent study showed no negative effect on cognition following smoking cessation, while it resulted in an improvement (Boggs et al., Reference Boggs, Surti, Esterlis, Pittman, Cosgrove, Sewell and D'Souza2018). In a large cohort study of 1094 individuals with SZ, smoking cessation improved speed processing (Vermeulen et al., Reference Vermeulen, Schirmbeck, Blankers, van Tricht, Bruggeman and van den Brink2018).
The fourth and last hypothesis is that tobacco is a precipitating factor of schizophrenic disorders. Tobacco initiation precedes onset of SZ in 90% of subjects, is associated with an earlier age at onset, and both light and heavy smoking are prospectively associated with a higher risk of developing SZ (Gage & Munafò, Reference Gage and Munafò2015; Gurillo, Jauhar, Murray, & MacCabe, Reference Gurillo, Jauhar, Murray and MacCabe2015; Kendler, Lönn, Sundquist, & Sundquist, Reference Kendler, Lönn, Sundquist and Sundquist2015). Following this hypothesis, if tobacco interferes with late neurodevelopment during adolescence as cannabis does (Mallet, Ramoz, Le Strat, Gorwood, & Dubertret, Reference Mallet, Ramoz, Le Strat, Gorwood and Dubertret2017b), then smokers may present less cognitive deficits because of different and low early neurodevelopmental load. In the same way, ex-smokers could also represent a distinct clinical and cognitive profile among non-smokers (with an exposition during late neurodevelopment and no current effect of tobacco on cognition). Ex-smokers would then present a different pathophysiological and neurodevelopmental pathway leading to the development of SZ, during late adolescence. It would thus concern only ex-smokers who used to smoke during late neurodevelopmental period. Yet, little is known on the differences concerning cognitive profile between smokers, ex-smokers and non-smokers, while it may increase knowledge on the underlying factors contributing to this association. More generally, those who have been able to quit smoking (ex-smokers) and those who never took up cigarette smoking (never-smokers) could represent two subtypes of patients, who could potentially provide us with important insights into the factors that provide some form of ‘protection’ from tobacco dependence, either at the initiation (never-smokers) or at relapse stage (former smokers). This hypothesis remains to be explored.
Our main objective is to compare the cognitive performance (IQ and cognitive processes) according to smoking status, in a large multicentre sample of well-stabilized individuals with SZ. We hypothesize that (i) smokers with SZ do not present better cognitive performances, arguing against the auto-medication hypothesis; and (ii) ex- and never-smokers represent two different neurocognitive profiles, underlied by two different neurodevelopmental and pathophysiological pathways to SZ.
Materials and methods
Study population
This study is a cross-sectional study based on a national cohort issued from FondaMental Academic Centers of Expertise (FACE) for Schizophrenia (SZ), the FACE-SZ cohort (Schürhoff et al., Reference Schürhoff, Fond, Berna, Bulzacka, Vilain and Capdevielle2015). The FACE-SZ cohort is based on a French national network of 10 Expert Centers. This network was set up by the Fondation FondaMental and funded by the French Ministries of Research and of Health. The cohort and the clinical and cognitive variables have been extensively described (Mallet et al., Reference Mallet, Godin, Le Strat, Mazer, Berna and Boyer2021; Schürhoff et al., Reference Schürhoff, Fond, Berna, Bulzacka, Vilain and Capdevielle2015). Persons are referred by their GP or their psychiatrist, sometimes at the beginning of the disease (to confirm the diagnosis) or sometimes for a therapeutical advice. Diagnoses were carried out by two psychiatrists according to the Structured Clinical Interview (SCID) for DSM Disorders (First, Spitzer, Gibbon, & Williams, Reference First, Spitzer, Gibbon and Williams2002).
All individuals included were outpatients, living outside the hospital and on stable medication for more than 4 weeks. The assessment protocol was approved by the relevant ethical review board (Comité de Protection des Personnes, CPP-Ile-de-France IX, 18 January 2010). All data were collected anonymously.
Data collected
Clinical and sociodemographic measures
All the collected data have recently been described (Mallet et al., Reference Mallet, Godin, Le Strat, Mazer, Berna and Boyer2021). Duration of untreated psychosis was recorded. It was calculated, using the age at onset and the age at the first treatment. Age at onset and age at first treatment were retrospectively determined during the SCID-based semi-structured interview with the patient and her/his relatives as well as by available medical records. This assessment took place during a comprehensive 1-day clinical evaluation. Age at first treatment was defined as the first time psychosis was ‘adequately’ treated, which is the most common definition (for review, see Register-Brown & Hong, Reference Register-Brown and Hong2014). Psychotic and general psychopathology was assessed using the PANSS (Positive and Negative Syndrome Scale) (Kay, Fiszbein, & Opler, Reference Kay, Fiszbein and Opler1987). Current depressive symptoms were evaluated with the Calgary Depressive Rating Scale (CDRS) (Bernard, Lançon, Auquier, Reine, & Addington, Reference Bernard, Lançon, Auquier, Reine and Addington1998). Manic symptoms were assessed with the YMRS (Young Mania Rating Scale) (Young, Biggs, Ziegler, & Meyer, Reference Young, Biggs, Ziegler and Meyer1978). Global functioning was evaluated with the GAF (Global Assessment of Functioning) (Endicott, Spitzer, Fleiss, & Cohen, Reference Endicott, Spitzer, Fleiss and Cohen1976). Ongoing antipsychotic, antidepressant or mood-stabilizer treatments were recorded. Chlorpromazine equivalent doses (CPZ100eq) were calculated (Leucht et al., Reference Leucht, Samara, Heres, Patel, Furukawa, Cipriani and Davis2015). Self-reported adherence to pharmacological treatment was evaluated with the MARS (Medication Adherence Rating Scale) (Misdrahi, Verdoux, Llorca, & Baylé, Reference Misdrahi, Verdoux, Llorca and Baylé2004).
Smoking status
Current smoking was defined as current daily smoking because non-daily smoking is rare in patients with SZ (de Leon & Diaz, Reference de Leon and Diaz2005). Participants who reported that they did not smoke or smoke daily were also considered non-smokers, as were participants who took a nicotine replacement or used electronic cigarettes. As non-smokers include ex-smoker, we also defined a group of ex-smokers to allow comparisons. Ex-smokers were defined according to the glossary of the US centers for diseases control and prevention (www.cdc.gov), as having smoked at least 100 cigarettes in his/her lifetime (but who had quit smoking at the time of interview).
Nicotine dependence was measured by the Fagerström Test for Nicotine Dependence (FTND) (Heatherton, Kozlowski, Frecker, & Fagerström, Reference Heatherton, Kozlowski, Frecker and Fagerström1991). Severe nicotine dependence was defined by a score ⩾7.
Neuropsychological measures
The standardized test batteries complied with the recommendations issued by the MATRICS™ Consensus Cognitive Battery (MCCB™) (Nuechterlein et al., Reference Nuechterlein, Green, Kern, Baade, Barch, Cohen and Marder2008).
Wechsler Adult Intelligence Scale (WAIS)-3rd Edition provides a measure of general intellectual function. The seven subtest short form was used and allowed exploration of the following cognitive areas: picture completion (visual exploration and detail perception), digit-symbol coding (visual-motor coordination, motor and mental speed), similarities (abstract verbal reasoning), arithmetic (mathematical problem solving), matrix reasoning (non-verbal abstract problem solving, inductive spatial reasoning), digit span (attention, working memory, mental control), information (general information acquired from culture, semantic memory). Intellectual functioning was evaluated with information and matrix reasoning subtests. An additional subtest, letter-number sequencing was administered, which along with two other primary subtests, digit span and arithmetic, allowed the calculation of a Working Memory Index (WMI; auditory working memory and mental control).
Barona Index is a demographically based regression method to estimate premorbid intelligence in terms of index scores on the WAIS.
The California Verbal Learning Test (CVLT) was used to evaluate verbal memory and executive functions (Delis, Kramer, Kaplan, & Ober, Reference Delis, Kramer, Kaplan and Ober2000).
The Trail Making Test (TMT) allowed assessment of executive functions as well as speed of processing and visual attention (Reitan, Reference Reitan1958).
Comorbidities
-
− Current and past addictions were systematically recorded. The assessment of addictions encompassed several screening questions to the participants: ‘Have you ever consumed alcohol of any kind more than 10 times in your life? (yes or no answer, primary abstainer? (no alcohol consumption at all in your lifetime?)/Have you ever used cannabis or hashish, joints, marijuana, or grass more than 10 times in your life? (yes/no/primary abstainer? (not a single lifetime use of cannabis?)/Have you ever used illicit drugs, or medicines not prescribed by a doctor, such as sedatives, benzodiazepines, or other medicines, crack, cocaine, amphetamines, heroin, subutex, methadone, hallucinogens, or other products, more than 10 times in your life?/Have you ever had to lie to people close to you about your gambling behaviour?/Have you ever felt the need to bet more and more money?’ After this initial screening, other more detailed questions were asked, according previous answers, to define the presence or absence of current/past addictions. Alcohol and/or cannabis dependence was defined according to a SCID 1.0-based semi-structured interview with the patient.
-
− Learning and/or language disorders were screened with medical childhood records, as well as family information. The information was reported by ‘yes/no’ if at least one of the following disorders was found: dyslexia, dysorthographia, dyscalculia, dysphasia, dyspraxia, language delay, stuttering, delay in walking.
Statistical analysis
All variables were presented using measures of means and dispersion (standard deviation) for continuous data and frequency distribution for categorical variables. First, we compared SZ current smoker's v. non-current smokers (including ex-current and never-smokers). Univariable associations between sociodemographics, clinical and treatments with tobacco status were performed using parametric or non-parametric tests (χ2test or fisher test for categorical variables and Student test or Mann–Whitney Wilcox tests for continuous variables after examination for normal distribution). All variables associated with tobacco smoking with a p value < 0.2 in univariable analyses were then included in a multivariable logistic regression model with tobacco smoking considered as dependent variable. In order to study the relationship between smoking status and neuropsychological characteristics, first we used a multivariable linear regression analysis (online Supplementary Table S1). Based on all neuropsychological variables significantly associated with smoking status with a p value < 0.20, and in order to eliminate multicollinearity associated with cognitive tests, we performed a principal component analysis (PCA) with varimax rotation to reduce the large number of cognitive tests to a smaller number of uncorrelated cognitive component scores. The number of components was chosen based on the Kaiser stopping criterion (i.e. all component eigenvalues >1) and the scree test. The use of component scores as the independent variables in multivariable models is considered as relevant in the case of multicollinearity (Tabachnick & Fidel, Reference Tabachnick and Fidel2001). To deal with missing values of cognitive variables, data were imputed for the PCA. Lastly, we performed a multivariable logistic regression analysis including tobacco status as a dependent variable, component scores as independent variable and a set of confounding factors selected either from the multivariable analysis of sociodemographic, clinical and treatment characteristics, with selection based on a threshold p value < 0.20 or on confounding factors known to impact neurocognitive variables and/or tobacco status (income and psychotropic medications). Finally, analyses were adjusted for: age, years of education, income, psychotic symptomatology, lifetime cannabis and alcohol use disorder, chlorpromazine equivalents (CPZeq tot), treatment adherence and psychotropic medication (benzodiazepine, anticholinergic and antidepressant).
In order to better explore if ex-smokers may represent a specific subgroup, comparison between the three groups of tobacco status (never-smokers, ex-smokers and never-smokers) were performed using the same method (univariable and multivariable logistic regression analysis). Neurocognitive components were thus compared between ex-smokers and never-smokers particularly adjusting for age, years of education, income, psychotic symptomatology, lifetime cannabis and alcohol use disorder, chlorpromazine equivalents (CPZeq tot), treatment adherence and psychotropic medication (benzodiazepine, anticholinergic and antidepressant). As recommended by Bender & Lange (Reference Bender and Lange2001), when necessary, univariate analyses have been corrected for multiple comparisons with Holm method, more powerful than the Bonferroni method (online Supplementary Table S2).
Results
A sample of 1233 community-dwelling stable SZ subjects enrolled in the FACE-SZ cohort was included in the study. Table 1 shows sociodemographic and clinical characteristics of the sample and the associations with tobacco smoking.
Table 1. Sociodemographic and clinical characteristics according to current tobacco status in a sample of n = 1233 SZ patients
s.d., standard deviation; DUP, duration of untreated psychosis; GAF, Global assessment of Functioning; PANSS, Positive and Negative Syndrome Scale; CDSS, Calgary Depression rating Scale for Schizophrenia; YMRS, Young Mania rating Scale; CPZeq, chlorpromazine equivalent doses (CPZ100eq), calculated according to the minimum effective dose method (Leucht et al., Reference Leucht, Samara, Heres, Patel, Furukawa, Cipriani and Davis2015); MARS, Medication Adherence Rating Scale.
a χ2 test for categorical variable and student test or Wilcox test for continuous variables. Logistic regression adjusted on sex, mean years of education, income, diagnosis, negative symptoms scale, Calgary depression rating scale, lifetime alcohol consumption disorder, lifetime cannabis consumption disorder, chlorpromazine equivalent, adherence rating scale and benzodiazepine.
Note: Statistically significant in bold (p < 0.05).
Subjects were mostly men (n = 967, 73.9%), mean aged 31.5 (±9.5) years old. Mean illness duration was 10.02 years (s.d. = 7.8). Overall, 662 patients were smokers (53.7%).
Association with current tobacco smoking
Sociodemographic and clinical correlates
Table 1 shows sociodemographic and clinical characteristics associated with tobacco smoking. Mean number pack years was 13.35 (s.d. = 13.2) in current smokers, with around one-third being nicotine-dependent (Fagerström score ⩾7, 33.7% of the smokers sample).
In the multivariable analysis, only male sex, less education, lifetime cannabis or alcohol use disorder and higher doses of antipsychotic treatment (CPZ equivalent) remained associated to current tobacco smoking.
Cognitive correlates of current smoking
Online Supplementary Table S1 shows cognitive correlates of current tobacco smoking.
After adjusting for all potential confounding factors, the multivariate analysis revealed lower performances in current smokers compared to non-smokers with a lower premorbid IQ (p < 0.001), current IQ (full scale, 7 subtest short form, p = 0.027), general knowledge (information, p = 0.002) and lower executive functioning (matrix reasoning, p = 0.027).
Results of the PCA
Three components emerged in the PCA (Table 2) that explained 71.6% of the variance. Each value of component was strictly superior to 1. The content of each component was considered as relevant and meaningful by two experts in neuropsychology (JM and CD) and named ‘general intellectual ability/abstract thinking’, ‘memory and learning’ and ‘working memory/processing speed’.
Table 2. Principal component analysis after varimax rotation, on neuropsychology's variables impute (tobacco defined by: no-smoker and ex-smoker v. smoker), n = 1309
Three main components emerged from the PCA. Each value of component is strictly superior to 1 and they explained 71.6% of the variance.
After a multivariable analysis (Table 3), current smoking remained associated with a lifetime alcohol use disorder (p = 0.026) and a lifetime cannabis use disorder (p < 0.001). At a cognitive level, current smoking was associated with worse performances at the component 1 ‘general intellectual ability/abstract thinking’ (adjusted OR 0.60, 95% IC 0.41–0.88, p = 0.010).
Table 3. Factors associated with current smoking (v. not) in a sample of n = 1233 subjects with schizophrenia
PANSS, Positive and Negative Syndrome Scale; CPZeq, chlorpromazine equivalent doses (CPZ100eq), calculated according to the minimum effective dose method (Leucht et al., Reference Leucht, Samara, Heres, Patel, Furukawa, Cipriani and Davis2015); MARS, Medication Adherence Rating Scale.
Note: Statistically significant in bold (p < 0.05).
The factors associated with tobacco history in non-smokers are detailed later (Table 4).
Table 4. Factors associated with tobacco history in non-smokers (ex- v. never-smokers) in a sample of n = 1233 subjects with schizophrenia
PANSS, Positive and Negative Syndrome Scale; CPZeq, chlorpromazine equivalent doses (CPZ100eq), calculated according to the minimum effective dose method (Leucht et al., Reference Leucht, Samara, Heres, Patel, Furukawa, Cipriani and Davis2015); MARS, Medication Adherence Rating Scale.
Note: Statistically significant in bold (p < 0.05).
Comparison of the three smoking status: current smokers, ex-smokers and never-smokers
Sociodemographic and clinical correlates
Characteristics of the three groups are resumed in online Supplementary Table S2 (two by two comparison of the three groups). This result allowed us to compare groups two by two, and particularly ex-smokers with no smokers.
Ex-smokers only constituted a minority of non-smokers (n = 117, 20.49%) and of the total sample (n = 117, 9.5%). Mean number of pack years was 9.29 (s.d. = 8.9) in ex-smokers and 13.35 (s.d. = 13.2) in current smokers. Age at first cigarette was lower in subjects still smoking in comparison to ex-smokers (m = 15.50 yo (s.d. = 4.1) v. m = 16.43 yo (s.d. = 4.1), p = 0.015).
• Two by two comparisons of the three groups (online Supplementary Table S2)
Female sex was more prevalent in the never-smoker group, in comparison to current smokers (p < 0.001), and there was no difference between never- and ex-smokers or ex- and current smokers. Ex-smokers were older than current and never-smokers (p < 0.001).
The three groups differed for the level of education when comparing 2 by 2, with current smokers being the less educated and ex-smokers the more (ex-smokers v. never, p = 0.020).
Age at first antipsychotic treatment globally differed (but no difference emerged when comparing groups 2 by 2), as illness duration (longer in ex-smokers, in line with their older age).
Regarding clinical characteristics, the groups differed for psychotic symptomatology (PANSS total score lower in ex-smokers), with a significant difference when comparing ex and never-smokers (more psychotic symptoms in never-smokers, p = 0.008) and only a trend when comparing with current smokers, p = 0.082). More specifically, they differed for negative symptoms (less negative symptoms in ex- and current smokers than in never-smokers: p < 0.001, but no difference between ex- and current smokers) and general psychopathology (PANSS general score, ex-smokers having a lower score than never-smokers: p = 0.040, but no difference with current smokers). We observed no difference regarding thymic symptoms (manic or depressive).
The three groups also differed for current addiction comorbidities. Current and lifetime substance use disorder and cannabis use disorder being associated with current smoking. Current alcohol use disorder was also associated with current smoking whereas lifetime alcohol use disorder was similar between ex- and current smokers. When comparing ex-smokers and never-smokers, the two groups differed with less current and lifetime use disorder in never-smokers.
When examining medication, current smoking was not associated with higher total dose of CPZeqtot, as only never-smokers had lower CPZeqtot in comparison to current smokers (p < 0.001), while ex-smokers and smokers did not significantly differ. Current smokers had a lower adherence to medication than both ex- and never-smokers (with no difference between ex- and never smokers).
– Global univariate and multivariable analysis of the three groups (data not shown)
In multivariable analysis (adjusted on variables significant at the 0.20 α level), the smoking status was associated to lifetime alcohol use disorder (p = 0.021), lifetime cannabis use disorder (p < 0.001) and the total dose of CPZeqtot (p = 0.003). The level of education was at the limit of the approached significance level (p = 0.054).
Cognitive correlates of smoking status: smokers, ex- and non-smokers
In the multivariable analyses (data not shown), the premorbid IQ and general knowledge task (information) remained significantly associated with the smoking status. Deeper analyses with 2 by 2 comparisons distinguish different profiles. When examining total premorbid total IQ, the three groups significantly differed with, from higher to lower premorbid total IQ: ex-smokers, never-smokers, current smokers (all p < 0.001). Regarding the general knowledge task, the three groups differ with, from higher to lower scores, never smokers, ex-smokers and current smokers (all p < 0.001).
When further analysing the non-smoking group, ex-smokers and never-smokers differed for age [ex-smokers being older, aOR 1.05 (95% CI 1.00–1.11, p = 0.047)], mean income [ex-smokers having higher incomes, aOR 3.39 (95% CI 1.17–10.3, p = 0.026)], a lifetime cannabis use disorder [aOR 33.2 (95% CI 10.1–135.1, p < 0.001)] and CPZeq doses [aOR 1.00 (95% CI 1.00–1.01, p = 0.005)] (Table 4). We observed no difference for the three cognitive components of the PCA, or for the level of education, but the component 2 ‘working memory/processing speed’ was at the limit of the approached significance level (less performances in ex-smokers than never-smokers).
Discussion
This large national sample of non-selected subjects with well-stabilized SZ aimed to compare the effect of smoking cigarettes on cognition. We found no cognitive advantage in smokers in comparison to non-smokers, especially in terms of executive function or attention. Conversely, smokers would rather present lower general intellectual abilities. Indeed, the ‘general intellectual ability/abstract reasoning’ cognitive factor was lower in smokers in comparison to non-smokers, independently of sex, education, income, age at onset, illness duration, psychotic symptomatology, alcohol or cannabis use disorder, medication doses, type of medication and adherence. Moreover, we found no major cognitive performance differences between ex-smokers and never-smokers in this sample (except higher premorbid IQ in ex-smokers compared to never or current smokers).
Smoking behaviour and SZ
When considering separately the sociodemographic and clinical characteristics of smokers, this study confirms previous findings (Mallet et al., Reference Mallet, Le Strat, Schürhoff, Mazer, Portalier and Andrianarisoa2017a, Reference Mallet, Ramoz, Le Strat, Gorwood and Dubertret2017b, Reference Mallet, Le Strat, Schürhoff, Mazer, Portalier and Andrianarisoa2018): tobacco smoking is common in SZ with a prevalence around 53% in this large sample. This proportion seems stable in this population while prevalence recently decreased in the French general population [25.5% in 2020 (Pasquereau, Reference Pasquereaus.d.)]. Around one-third (33.7%) of this sample of smokers with SZ presented a severe nicotine dependence, which also aligns with previous studies (Yee et al., Reference Yee, Bt Nek Mohamed, Binti Hashim, Loh, Harbajan Singh, Ng and Jambunathan2015) and is higher than in smokers from the general population. Smoking is associated with male sex, a lower education level, lifetime cannabis or alcohol use disorder (as in the general population) and higher doses of antipsychotic treatment (CPZ equivalent) but no significant association between tobacco smoking and age of illness onset, depression, adverse childhood experience, history of suicidal behaviour, global functioning. Smokers had a higher medication dosage (CPZeqtot) but groups did not differ regarding treatment observance. When considering the whole cognitive and clinical characteristics in a multivariable analyse (PCA), we found that current tobacco smoking was strongly associated with less general intellectual ability/abstract thinking (component 1 of the PCA), and a history of cannabis or alcohol use disorder.
Comparison of ex- v. never-smokers
Interestingly, when splitting the non-smoking sample in two groups, namely ex- v. never-smokers, various characteristics emerged in the 2 by 2 comparisons after correction for multiple comparisons. Ex-smokers were older and had a longer illness duration (as expected), and were more educated and with a higher income than never-smokers. At a clinical level, they presented less total psychotic symptoms, in particular, less negative and general symptoms. Ex-smokers were more likely to present a current or past substance use disorder. At a cognitive level, only the premorbid IQ differed (more elevated in ex- than never-smokers). When analysing data in the PCA using three cognitive components, after adjustment on all confounding variables, the ex- and never-smokers only differed for age, outcome, history of cannabis use disorder (ex-smokers suffering more often of a history of cannabis use disorder). The latter is often associated with tobacco smoking, with a common shared genetic substrate (dual users, particularly in subjects with mental health problems) (Hindocha, Brose, Walsh, & Cheeseman, Reference Hindocha, Brose, Walsh and Cheeseman2021). The global cognitive results do not support previous findings of two earlier studies with small sample size from the same team in which the never-smoker group had the poorest performance on neuropsychological tests assessing attention and processing speed (Wing, Bacher, Sacco, & George, Reference Wing, Bacher, Sacco and George2011). However, the initial and more subtle comparisons analysing each subtask 2 by 2 of the two groups could be in favour of a less neurodevelopmental load in the ex-smokers group compared to never-smokers, as they presented the highest premorbid IQ, achieved a higher education level and presented less negative symptoms.
Cigarettes contain numerous neurotoxic substances, making the net effect of cigarette smoking on brain function and structure complex. Recently, authors found negative additive effects of SZ diagnosis and smoking status on the brain grey matter (in the left prefrontal cortex) (Yokoyama et al., Reference Yokoyama, Sasaki, Mori, Ono, Tsurumi, Kawada and Takahashi2018). Interestingly, the decrease in the grey matter volume was associated with greater numbers of cigarette pack years and severe positive and negative symptoms. Another recent study by Schneider et al. reported smaller hippocampal and prefrontal volumes in smoking SZ patients compared to non-smoking patients (Schneider et al., Reference Schneider, White, Hass, Geisler, Wallace, Roessner and Ehrlich2014). These neuroimaging results are in opposition with a previous study supporting the self-medication hypothesis and hypothesizing that tobacco preserved decreased grey matter in SZ (Tregellas et al., Reference Tregellas, Shatti, Tanabe, Martin, Gibson, Wylie and Rojas2007). Moreover, they also suggest that tobacco consumption can negatively interfere with late neurodevelopment and impact cognition in future ex-smokers, leading to a subgroup with a distinct clinical and cognitive profile in non-smoking patients (better premorbid IQ and education than never and current smokers, contrasting with a final equivalent cognitive impairment, comparable in ex- and never-smokers). Nicotine or tobacco can interfere with late neurodevelopment through direct or indirect action [inflammatory mechanisms, hypersensitization of D2 receptors (Novak, Seeman, & Le Foll, Reference Novak, Seeman and Le Foll2010)]. These effects could precipitate a schizophrenic disorder in vulnerable subjects, according to the neurodevelopmental and multifactorial approaches of the disease (Howes et al., Reference Howes, Kambeitz, Kim, Stahl, Slifstein, Abi-Dargham and Kapur2012). This paradigm is corroborated by recent findings on the possible role of tobacco consumption on SZ onset (Gage & Munafò, Reference Gage and Munafò2015; Gurillo et al., Reference Gurillo, Jauhar, Murray and MacCabe2015; Kendler et al., Reference Kendler, Lönn, Sundquist and Sundquist2015).
Overall, future clinical studies should distinguish ex- and never-smokers, as these two groups may present different neurodevelopmental pathways from childhood to adulthood.
Current tobacco use
Current findings do not support an attempt to self-medicate executive or attention dysfunction with tobacco. In particular, the slight positive effects of current smoking on attention in the trail-making test A in the univariate analysis were no longer relevant in the multivariable analysis. In addition, the digit symbol coding test (measuring the same cognitive function, namely the speed of processing) showed opposite results in univariate analysis that also disappeared in multivariable analysis. Moreover, current smokers showed less performance in the matrix reasoning task (executive functions/problem solving). Thus, looking at other more experimental studies on the effects of smoking/nicotine on attention disorders in SZ patients, it seems unlikely that this slight positive effect of tobacco on attention (not sustaining adjustment on confounding factors and only found in slight working memory load tasks as TMT A, but not in the digit symbol coding test) makes patients keep smoking. Overall, our results are in line with those of a recent study that showed a deterioration in the speed of processing in smoking patients, albeit with another methodology [use of the WAIS only(Vermeulen et al., Reference Vermeulen, Schirmbeck, Blankers, van Tricht, Bruggeman and van den Brink2018)].
Contrary to what is supported by the self-medication hypothesis, the whole PCA multivariable analysis demonstrated that after adjustment on confounding factors, current smokers were clearly less performant in the general intellectual ability (abstract thinking) and showed no advantage on the second component (working memory/ processing speed). They also were more likely to present a history of cannabis use or alcohol use disorder.
Further experimental and longitudinal studies are warranted to understand the neurocognitive impact of smoking in SZ. Indeed, recent fMRI findings showed that smoking in SZ might fail as an attempt of normalizing cognitive function (‘recovery effect’ on cognition) but could prevent worsening of previously impaired cognitive function (‘preservation effect’ on cognition) (Liao et al., Reference Liao, Fan, Yang, Li, Duan, Cui and Chen2019). In this small study, cigarette smoking could elicit a distinct improvement in the dynamics of triple networks in SZ but not in healthy controls. It would partly support the self-medication hypothesis in terms of brain functional dynamic.
Ex-smokers and current smokers
At a clinical level, the two by two comparison showed that ex-smokers were older, had a higher education level, a higher outcome and had a longer illness duration compared to current smokers. They also presented less current comorbid addiction. Finally, they were more adherent to medication (MARS score). At a cognitive level, ex-smokers present a higher premorbid IQ and a higher education level, which might have protected them against pursuing smoking. However, no current cognitive difference was observed at a distinct (2 by 2 comparisons for subtasks) level.
Strengths and limits
The main strengths of this study include the use of homogenous and exhaustive standardized protocols and neuropsychological assessment in a large national multicentric study, which provides a ‘real-world’ patients sample. This is the largest study to date, exploring the cognitive correlates of smoking status in SZ. We therefore included many clinical and therapeutic variables in the multivariate analyses. We also had the statistical power to distinguish ex-, current and never-smokers. Moreover, as smoking is associated with socio-economic status, that is associated with education level, we had the opportunity to disentangle these biases on cognitive results by adjusting on economic outcome and education levels. However, the study has several limitations. First, this study is cross-sectional; therefore, we were not able to investigate the direction of effect between tobacco smoking and SZ, symptoms, cognition or learning/language disorders. Longitudinal studies would be valuable to investigate the temporal relationship between these factors. Second, our out-patients sample is not representative of all patients with SZ, particularly because institutionalized, hospitalized or highly disabled patients (making the cognitive assessment difficult) were not referred to the Expert Centres. The FACE-SZ cohort is not an epidemiological cohort but a cohort of persons attending secondary or tertiary care centres. However, the multicentric network of expert centres appears as a strength to counterbalance this bias, covering the whole French territory and leading to homogenous groups of patients in terms of clinical severity and exposure to tobacco. Only samples selected for the purpose of epidemiological study can warrant a sufficient level of representativeness. Moreover, in-patients present biased neuropsychological examinations. Third, the age at tobacco smoking onset was not considered. Future studies should explore if tobacco smoking before SZ onset has a different impact than after SZ onset. It should also be determined if heavy smokers have different characteristics compared to other SZ-smokers. Finally, our study considers that participants smoking non-daily or e-cigarette users as non-smokers whereas they are also exposed to nicotine. Similarly, a specific category of cannabis only (non-cigarette) smokers was not individualized in this study because this use of cannabis is very rare in France, while cannabis joints contain tobacco/nicotine and may account for ‘tobacco smoker status’.
This is due to the design of the FACE-SZ cohort that does not allow to consider this characteristic. Future studies should examine this distinct pattern of exposure.
Conclusion
The prevalence of current tobacco smoking in French community-dwelling SZ subjects is still more than twice higher than in the general population, with a high level of nicotine dependence. This is the largest study to date that provides strong evidence that chronic smoking is associated with cognitive impairment in SZ, arguing against the self-medication hypothesis as a contributor to the high prevalence of smoking in SZ. The present study also suggests a distinct clinical and cognitive profile in ex- v. never-smokers. Pre-clinical and longitudinal studies are warranted to determine the developmental impact of tobacco on neurocognition.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291722002574
Data
The data that support the findings of this study are available from Foundation Fondamental. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the corresponding author with the permission of Fondation Fondamental.
Acknowledgements
Authors would like to thank all participants and their families.
Author contributions
All authors acquired the data, which J. Mallet, Y. Dansou, Y. Le Strat and C. Dubertret analysed. O.Godin completed the statistical analyses. J. Mallet wrote the first draft, which all authors reviewed. All authors approved the final version to be published and can certify that no other individuals not listed as authors have made substantial contributions to the paper.
Financial support
This work was funded by AP-HP (Assistance Publique des Hôpitaux de Paris), Fondation FondaMental (RTRS Santé Mentale), by the ‘Investissements d'Avenir’ programmes managed by the Agence Nationale de Recherche (ANR), France under references ANR-11-IDEX-0004-02 and ANR-10-COHO-10-01, and by INSERM (Institut National de la Santé et de la Recherche Médicale).
Conflict of interest
The authors report no financial or other relationship relevant to the subject of this article.
Ethical standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
Consent to participate
All persons gave their informed consent prior to their inclusion in the study.
