CHAPITRE 2 Contexte et résumé de l’étude

Table des matières

Les données de la présente étude ont été obtenues à partir d’une étude de cohorte visant à déterminer si l’exposition à une demande psychologique élevée et à une latitude décisionnelle faible augmentait le risque de souffrir d’une maladie cardio-vasculaire. Le groupe de référence était composé des gens non exposés à ces deux contraintes. Il y eut deux collectes de données. La collecte initiale a eu lieu de 1991 à 1993 et la collecte des données de suivi a eu lieu de 1999 à 2003. La cohorte était composée de cols blancs de la région métropolitaine de la ville de Québec.

Un article sur les données et les résultats de la collecte de base de l’étude (1991 – 1993) a été publié en 1999 (19). L’objectif principal de cette étude transversale était de mesurer l’association entre une exposition simultanée à une forte demande psychologique (DP+) et à une faible latitude décisionnelle (LD-) au travail et la consommation de médicaments psychotropes. L’étude, effectuée auprès de 2786 cols blancs de la région métropolitaine de la ville de Québec, a trouvé une prévalence de consommation de 3,9%. La proportion totale d’exposés à DP+ et LD- était de 20,5%. L’association entre la tension au travail et l’usage de médicaments psychotropes était statistiquement significative (RC : 4,5 ; IC à 95% = 1,7 – 12,2) après ajustement pour le soutien social au travail, l’âge, le sexe, l’éducation, le revenu familial, le statut d’emploi, le type d’emploi, les événements stressants, le tabagisme, la consommation d’alcool et le niveau d’activité physique. Les auteurs en concluent que la tension au travail est un facteur déterminant en ce qui concerne l’usage de médicaments psychotropes, mais que le soutien social ne semble pas modifier cette association.

Essentiellement, la présente analyse a porté sur les données recueillies en 1999 – 2003 en utilisant les mêmes variables et analyses qu’à la collecte des données de base (1991 – 1993). Toutefois, quelques variables supplémentaires et la modélisation à l’aide du modèle d’inadéquation efforts extrinsèques-reconnaissance de Siegrist ont été ajoutées afin d’en vérifier la pertinence dans la présente relation. Cette analyse transversale permettra de voir s’il y a eu changement dans la prévalence et l’origine de l’utilisation des médicaments psychotropes au cours de la dernière décennie dans notre population à l’étude.

Tout comme à la collecte des données de base, les renseignements ont été recueillis à l’aide de questionnaires auto administrés recensant la demande psychologique et la latitude décisionnelle au travail, les efforts extrinsèques et intrinsèques, la reconnaissance, la consommation de médicaments psychotropes au cours des dernières 48 heures et une série de caractéristiques sociodémographiques. La population admissible était constituée de travailleurs provenant de plusieurs organismes gouvernementaux et d’une entreprise privée de la région métropolitaine de la ville de Québec travaillant à temps plein (35 heures et plus). De plus, les participants admissibles devaient avoir occupé leur poste actuel pour une période minimale de 12 mois consécutifs. Les questionnaires furent remplis sur les lieux de travail et les employés se voyaient libérés de leurs tâches pendant la période requise pour la complétion du questionnaire dans des lieux spécialement aménagés à cet effet.

I) Les participants à la collecte des données de base (1991-1993) ont signé un formulaire de consentement (Annexe C). En acceptant les conditions énoncées, les participants acceptaient que l’équipe de recherche puisse les contacter ultérieurement. Le protocole de recherche a été approuvé par le comité d’éthique de l’hôpital du Saint-Sacrement.

II) Les participants à la collecte des données de suivi (1999-2003) ont signé un formulaire de consentement (Annexe C). Le protocole de recherche a été approuvé par le comité d’éthique de l’hôpital du Saint-Sacrement.

III) La confidentialité des données a été assurée par l’assignation d’un numéro d’identification à chaque participant. Aucune identification personnelle n’apparaît au fichier principal des données. Un fichier séparé contient l’information nominale des participants et leur numéro d’identification. Ce fichier n’est accessible qu’aux chercheurs principaux et au personnel de recherche assigné à la gestion des informations nominales.

Plusieurs auteurs ont contribué à l’élaboration du projet et de l’article. La présente section énumère les chercheurs ayant participé au projet et donne aussi une brève description de la contribution de chaque membre de l’équipe de recherche ainsi que leurs affiliations.

- Patrice Chartrand1,2 : Étudiant à la maîtrise; revue de littérature, analyses statistiques et rédaction de l’article.

- Jocelyne Moisan1,2 : Co-chercheur, directrice de recherche de Patrice Chartrand; planification du projet, analyse et interprétation des données sur les médicaments psychotropes, supervision des analyses et révision de l’article.

- Renée Bourbonnais3 : Co-chercheur; planification du projet, commentaires et suggestions sur l’article.

- Chantal Brisson1,3 : Chercheur principale de l’étude sur la tension au travail et le risque de maladies cardiovasculaires; planification du projet, commentaires et suggestions sur l’article.

- Michel Vézina3 : Co-chercheur; planification du projet, commentaires et suggestions sur l’article.

- Jean-Pierre Grégoire1,2 : Co-chercheur; analyse et interprétation des données sur les médicaments psychotropes.

Affiliations : 1 Unité de recherche en santé des populations (URESP).

2 Faculté de pharmacie de l’Université Laval.

3 Faculté de médecine de l’Université Laval.

3.

CHAPITRE 3 Article

Job strain, efforts - rewards imbalance and

psychotropic drug use

Résumé

Objectifs : Les objectifs de cette étude étaient de mesurer l’association entre la tension au travail ou le ratio efforts extrinsèques/reconnaissance et l’utilisation de médicaments psychotropes chez un groupe de cols blancs de la région de Québec. Le troisième objectif consistait à déterminer si les efforts intrinsèques modifiaient la seconde association. Méthodes : Les données sur les variables d’exposition, la consommation de médicaments psychotropes des dernières 48 heures et les caractéristiques sociodémographiques furent recueillies par questionnaires auto administrés. Chaque participant fut classé selon le niveau de tension au travail et selon l’adéquation ou l’inadéquation du ratio efforts extrinsèques/reconnaissance. Les rapports de cotes (RC) et les intervalles de confiance (IC) obtenus furent standardisés en fonction des caractéristiques sociodémographiques. Résultats : La présente étude comptait 5757 participants. Un RC de 1,5 (IC 95% 1,0-2,2) a été observé chez le groupe le plus exposé à la tension au travail et un RC de 1,7 (IC 95% 1,4-2,2) chez le groupe exposé à une inadéquation du ratio efforts extrinsèques/ reconnaissance. Conclusion : L’inadéquation efforts extrinsèques/reconnaissance semble reliée à la consommation de médicaments psychotropes. Les efforts intrinsèques, le soutien social et le sexe ne modifient pas ces associations.

Aims: The aims of this study were to examine the association between job strain or extrinsic efforts-rewards ratio and psychotropic drug use in a group of white-collar workers from Quebec City. The third aim was to determine the modifying effects of intrinsic efforts in the latter relation. Methods: Participants completed a questionnaire assessing demands and latitude at work, extrinsic and intrinsic efforts, rewards, psychoactive drugs consumption during the two days prior to data collection and sociodemographic variables. Participants were divided in four different job strain groups and also, as having a balanced or imbalanced extrinsic efforts-rewards ratio. ORs were adjusted for numerous covariables. Results: This study accounted 5757 participants. For job strain, adjusted OR for the most exposed group was 1.5 (95% CI 1.0-2.2). The imbalanced extrinsic efforts-reward group had an adjusted OR of 1.7 (95% CI 1.4-2.2). Conclusion: We found a positive association between an imbalanced extrinsic efforts-rewards ratio and psychotropic drug use. Intrinsic efforts, social support and gender did not modify the observed associations.

The frequent use of psychotropic drugs is associated with health problems like ataxia, confusion, depression, dizziness, drowsiness and memory loss (67). These problems can affect productivity and increase the risk of accident at work (36, 46, 68-71). The prevalence of psychotropic drug use varies from 3.5 to 22.0% for the whole population (2-12) and from 3.9 to 19.5% for workers (13-20). The prevalence of psychotropic drug use is higher among people of low socioeconomic status (32, 43), among women (3, 6-8, 10), the elderly (3, 6-8, 19) and people with low physical or psychological well-being (8, 10, 32). Because exposure factors are, for the most part controllable, modifiable or adjustable, a better understanding of their short and long-term effects could lead the way to a better organization and division of work to prevent psychotropic drug consumption. Such knowledge could also serve to better diagnose and prevent deleterious effects of psychotropic drug consumption among employees. The initiation of preventive measures or screening programs capable of effectively identifying more susceptible persons, deleterious behavior or problematic job conditions or positions would allow employers and employees to better manage their work life and therefore, job strain.

Since the beginning of the 80’s, many studies have focused on the association between job stressors or job strain and drug use (psychotropic drugs in particular) (13, 15-17, 19, 22, 33, 34, 37, 72). Most of these found an association between job strain and psychotropic drug use. In 1979, Robert Karasek put forward the demands-control model (23). This model has been validated and widely applied to either mental or physical health problems (73). It is based on the hypothesis that high psychological demands combined with low decision latitude produce mental strain that may have harmful effects on health (60). Psychological demands encompasses workload, that is, the quantity and pacing of work, time constraints, the level of concentration required and the frequency of interruptions. Decision latitude covers two factors: skill discretion and decision-making power. Skill discretion refers to the opportunity to use and develop one’s abilities and to be creative in one’s job. Decision latitude refers to the individual’s capacity to make decisions regarding his or her job. Moreover, social support at work from colleagues and supervisors might reduce the effect of job strain on mental health (60).

In the mid 90’s, Johannes Siegrist suggested a new approach to quantify conditions at work that could lead to illnesses. He theorized that an imbalance between extrinsic efforts and rewards was particularly stressful and so, deleterious to health when compared to a balance between those two factors. Extrinsic efforts are evaluated using six items: time pressure, interruptions, responsibilities, pressure to work overtime, physical demands and increasing demands. Occupational rewards are composed of the evaluation of three dimensions: esteem reward, monetary gratification and status control.

Siegrist also suggested that overcommitment, also called intrinsic efforts, could modify the association between extrinsic efforts and rewards. Intrinsic efforts are composed of four dimensions: need for approval, competitiveness, disproportionate irritability and inability to withdraw from work. Since then, several research teams worked on his theories and results obtained on cardiovascular diseases, psychiatric disorders and employee well-being are very promising (28, 64, 65). Siegrist’s model uses personality traits and coping characteristics not covered by Karasek’s model (27, 74). For the first time, we decided to use that model in order to study the relation between job conditions and psychotropic drug use.

The first aim of this study was to measure the association between a combined exposition to high psychological demands, low decision latitude at work and psychotropic drug use among white-collar workers. Participants we recruited were part of a cohort on cardiovascular health followed from 1991 to 2003. Using follow-up data (1999-2003), we conducted analyses very similar to what was done at baseline (1991-1993) to see if we would be able to reproduce results obtained at that time. Results of the cross-sectional study done with baseline data has been published elsewhere (19).

The second aim was to determine if an imbalance between extrinsic efforts and rewards was increasing the risk of using psychotropic drugs and if intrinsic efforts were modifying this relation.

The third aim was to determine if social support at work, social support outside of work and gender modified these associations.

All questionnaires were self-administered. Participants were met during working hours and were still remunerated by their employer during that period. All employers allowed time and adequate space to facilitate participation of their employees to the study.

Karasek’s Job Content Questionnaire, an 18-item self-administered questionnaire was used to measure psychological demands (9 items) and decision latitude (9 items) (76). The validity of the English version of this questionnaire is well documented (60). The validity of the French version used here has been established (77). Cronbach’s α coefficient, which measures the internal consistency of the questionnaire, was .83 for decision latitude and .73 for psychological demands.

For decision latitude, the total score ranged from 24 to 96, a higher score meaning higher decision latitude. The total score of psychological demands ranged from 9 to 36, a higher score meaning higher psychological demands. Both psychological demands and decision were dichotomized at the median value of the score distribution obtained for these indices in a probability sample of all workers in the province of Quebec (78). Participants were categorized as having high or low psychological demands and decision latitude according to the value of their scores. Quebec workers median score is 72 for decision latitude and 24 for psychological demands. The combination of these two variables formed four exposure groups that defined job strain. Workers exposed to high psychological demands and low decision latitude were included in the high-exposure group, while workers exposed to low psychological demands and high decision latitude constituted the reference group. The other two possible combinations formed the intermediate-exposure groups.

Siegrist’s efforts-rewards questionnaire is divided in three parts. The first part concerns extrinsic efforts. Originally, it was composed of six questions about time pressure, work interruptions, responsibility, pressure to work overtime, physical demands and increasing demands and is very similar to Karasek’s psychological demands questionnaire. Since questions about extrinsic efforts were not included in our study questionnaire, we used Karasek’s psychological demands as a proxy measure to estimate extrinsic efforts as it was done in the Whitehall prospective study on mental health disorders (65). The second part concerns rewards. Originally, it was composed of 11 questions concerning esteem (self-esteem and esteem from colleagues and superiors), monetary gratification and status control. In our study, we used nine of these questions to quantify rewards as was done in a study on cardiovascular diseases by Niedhammer and Siegrist (27). The third part concerns intrinsic efforts. Originally, it was composed of 29 items concerning the need for approval, competitiveness, disproportionate irritability and the inability to withdraw from work. In our study, we used 12 of these questions to quantify intrinsic efforts as it was done in the same study we referred to for the measurement of rewards (27).

For each question concerning rewards, a score of 1 or 2 points was assigned according to the answer given by the participant. A score of 1 meant a low level and a score of 2 meant a high level for every specific item. The rewards score range was 9 to 18. Since the number of questions differed from the original suggested questionnaire, we adjusted the weight of each scale to ensure correct computation of the extrinsic efforts (psychological demands)-rewards ratio. Participants were categorized as having a balanced (score of 1 or less) or imbalanced (score over 1) extrinsic efforts-reward ratio. For intrinsic efforts, the range of the score in our study was 12 to 48. We classified participants in the highest tertile (adjusted score of 33/48 and over, representing results obtained on a sample of workers in the province of Quebec) as having a high level of intrinsic efforts and the other two lowest tertiles as having a low level of intrinsic efforts.

Analyses were conducted using the SAS software package, version 8.01 (80). First, a descriptive analysis was carried out to determine the prevalence of psychotropic drug use during the previous two days according to the various study cofactors. Second, bivariate odds ratio (OR) and their 95% confidence intervals (CI) were calculated to measure the association between each cofactor and psychotropic drug use. Third, modification of the main relation between job strain and psychotropic drug use by social support at work, social support off work and gender was verified using the interaction terms method (81). The method was also applied to test the modification of the relation between the extrinsic efforts-rewards ratio and psychotropic drug use by intrinsic efforts, social support at work, social support outside of work and gender. Finally, a multivariate logistic regression model was used to compute the ORs adjusted for all cofactors used in baseline data analyses for comparability purposes (19). Also, a few others cofactors, significantly associated with psychotropic drug use were added for exploratory purposes. Multicolinearity was checked using a procedure described by Belsley, Kuh, and Welsch (1980), which is available in the REG procedure of the SAS software (82).

In all, 9191 individuals were included in the cardiovascular study cohort (Figure 1). Of those, 118 died and 1021 retired between baseline and follow-up analyses. 652 persons were part time employees and 456 did not cumulate at least one year at their present position. Among the 6944 eligible participants, 235 were lost to follow-up, 741 refused to participate, 75 had missing values on employment status and 136 persons had missing values on exposure variables. There were 5757 participants at follow-up representing a participation rate of 82.9%. Because of an extra 20 participants with missing values on efforts-rewards variables, the analyses for this model accounted 5737 participants.

In the two days preceding the participation to the study, 357 (6.2%) participants used at least one psychotropic drug. Of these, 201 (44.7%) used antidepressants, 173 (38.4%) used anxiolytics, 52 (11.6%) used hypnotics or sedatives, 15 (3.3%) used antipsychotics and 7 (2.0%) used psychostimulants. A total of 74 (1.3%) participants reported the use of drugs belonging to more than one class of psychotropic drug.

Among all study subjects, 1008 (17.5%) were exposed to the highest job strain group (high psychological demands (PD+) combined with low decision latitude (DL-)), 3680 (63.9%) were exposed to one of the two intermediate exposure group (PD+, DL+ or PD-, DL-) and 1069 (18.6%) were exposed to neither of the two constraints (PD-, DL+).

The crude OR for the association between exposure to high psychological demands combined with low decision latitude (job strain) and psychotropic drug use was 2.0 (95% CI 1.4-2.9). We observed an identical OR of 1.3 (95% CI 0.9-1.8) for each of the two intermediate levels. The adjusted ORs for the association between job strain and psychotropic drug use was 1.5 (95% CI 1.0-2.2). Adjustment was accounting for social support at and outside of work, age, gender, education, family income, employment status, occupation, stress associated with events that occurred during the last 12 months, practice of some meditation or relaxation technique, perception of health, cigarette smoking, alcohol consumption and physical activity (Table 1).

No modifying effect was noted with gender, social support at or outside of work and the relation between job strain and psychotropic drug use (Data not shown).

Social support at or outside of work were inversely related to psychotropic drug consumption. Our results showed a tendency toward a protective effect for those who had a low social support at work (OR 0.7, 95% CI 0.6-0.9) or outside of work (OR 0.6, 95% CI 0.5-0.7) (Table 1).

In our first analyses, we included all the legal psychotropic drugs that can be legally prescribed. Many of these drugs are not specifically designed to relieve stress but to treat medical conditions that can include stress among their symptoms (e.g.: antipsychotics or psychostimulants). Therefore, we attempted another series of analyses using only benzodiazepines (Table 2) and antidepressants (Table 3). These subclasses of drugs are mostly prescribed for stress management and are often used by researchers in studies similar to ours. For benzodiazepine use, the subgroup most exposed to job strain (PD+ DL-) had an adjusted OR of 1.3 (95% CI 0.8-2.3). For antidepressant use, the same subgroup had an adjusted OR of 1.4 (95% CI 0.9-2.4). There was a strong moderating effect for women using antidepressants when compared to men in the Karasek’s job strain model giving an OR of 1.8 (95% CI 1.2-2.6). Variables used for adjustment were the same as for the main analysis with all psychotropic drugs.

Karasek and Theorell demonstrated that psychological demands (PD) and decision latitude (DL) are interacting with one another (60). Therefore, they recommend using the combination of the two variables when describing job strain. Since we observed no association between job strain and psychotropic drug consumption, we attempted to consider psychological demand and decision latitude as separate variables to see if one dimension of job strain could better explain our results. The results suggest a barely significant effect of high psychological demands on psychotropic drug consumption with an OR of 1.4 (95% CI 1.1-1.8) and a non-significant relation when considering decision latitude alone with an OR of 1.1 (95% CI 0.8-1.4) (Table 4). These ORs were also adjusted for the same variables as the other analyses done in this study.

Among all study subjects, 1443 (25.1%) were exposed to an imbalance between extrinsic efforts and rewards. The crude OR for the association between an imbalanced extrinsic efforts-rewards ratio and psychotropic drug use was 1.9 (95% CI 1.5-2.4).

The adjusted OR for the association between an imbalanced extrinsic efforts-rewards ratio and psychotropic drug use was 1.7 (95% CI 1.4-2.2). The adjusted OR for the association between low rewards and psychotropic drug use was 1.8 (95% CI 1.4-2.4). Adjustment was accounting for social support at and outside of work, age, gender, education, family income, employment status, occupation, stress associated with events that occurred during the last 12 months, practice of some meditation or relaxation technique, perception of health, cigarette smoking, alcohol consumption and physical activity (Tables 5 and 7).

Also, 3428 (59.5%) participants were reporting a low level of rewards. The adjusted OR for the association between low rewards and psychotropic drug use was 1.8 (95% CI 1.4-2.4) (Table 8).

No interaction was noted between gender, social support at work, social support outside of work or intrinsic efforts and the relation between extrinsic efforts-rewards ratio and psychotropic drug use (Tables 11, 12 and 13).

Social support at or outside of work was inversely related to psychotropic drug consumption. Our results showed a tendency toward a protective effect for those who had a low social support at work (OR 0.7, 95% CI 0.6-0.9) or outside of work (OR 0.6, 95% CI 0.4-0.7) (Table 5).

We also examined the relation between extrinsic efforts-rewards and benzodiazepines use and between extrinsic efforts-rewards and antidepressant use. For benzodiazepine use, the subgroup exposed to an imbalance between extrinsic efforts-rewards had an adjusted OR of 1.9 (95% CI 1.4-2.7) (Table 9). For antidepressant use, the same subgroup had an adjusted OR of 1.5 (95% CI 1.1-2.2) (Table 7). There was a strong moderating effect for women using antidepressants when compared to men in the Siegrist’s extrinsic efforts-rewards model giving an OR of 1.9 (95% CI 1.3-2.8) (Table 7). Adjusted OR for the association between low rewards and benzodiazepines use and antidepressants use were respectively 2.4 (95% CI 1.6-3.5) and 1.9 (95% CI 1.3-2.7) (Tables 9 and 10). Variables used for adjustment were the same as for the main analysis with all psychotropic drugs.

Another analysis was done to examine if intrinsic efforts were modifying the relation between an imbalance in extrinsic efforts and rewards. We dichotomized the groups as showing a high level of intrinsic efforts (highest tertile) or a low level. We did the exercise for all psychotropic drugs, benzodiazepines alone and antidepressant alone. The following results always concern the imbalanced extrinsic efforts-rewards subgroups. The grouped psychotropic drugs analysis gave an OR of 1.9 (95% CI 1.3-2.7) for low intrinsic efforts and an OR of 1.4 (95% CI 1.0-2.1) for high intrinsic efforts. The benzodiazepines alone analysis gave an OR of 1.7 (95% CI 1.0-2.8) for low intrinsic efforts and an OR of 2.0 (95% CI 1.2-3.4) for high intrinsic efforts. The antidepressant alone analysis gave an OR of 1.9 (95% CI 1.2-2.9) for low intrinsic efforts and an OR of 1.2 (95% CI 0.7-2.0) for high intrinsic efforts. All analyses on specific drugs were adjusted for the same variables as the main analyses conducted with all psychotropic drugs (Tables 11, 12 and 13).

The results of the study presented here suggest that job strain is not associated with psychotropic drug use among white-collar workers, but an imbalance between extrinsic efforts and rewards is associated with psychotropic drug use. Furthermore, social support at and outside of work as well as gender do not appear to modify these associations. In addition, intrinsic efforts do not appear to modify the association between the extrinsic efforts and rewards ratio and the consumption of psychotropic drugs.

The prevalence of psychotropic drug use observed, 4.6% for men and 8.0% for women (undifferentiated results = 6.2%), are consistent with results obtained from recent studies using a 48 hours reference period. For example, in 1998, 7.7% of men and 6.5% of women representative of the Quebec population aged 15 years old or over reported use of a psychotropic drug within the last 48 hours (41). In 1992-93, a former Santé-Québec survey observed a prevalence of tranquilliser, sedative and sleeping pills use of 4.5% among adults aged 15 years old and over (83).

Factors found to be associated with psychotropic drug use in our study are consistent with those found in the literature. They include age, gender (for antidepressant analyses), cigarette smoking, stressful events that occurred during the past 12 months and perception of health. Some factors usually associated with psychotropic drug use in previous studies were not found significant in our study. Association with education level, family income, employment status, occupation, alcohol consumption and physical activity were all non- significant in adjusted analysis. Results suggest that these cofactors are not sufficient to induce psychotropic drug consumption. Also, they are not modifying the relation between job strain and psychotropic drug consumption.

Practice of a meditation/relaxation technique had a significant protective effect. The strong relationship between this activity and psychotropic drug consumption could be explained by a better ability to manage stress when practicing such techniques. In 2001, Van Der Klink and al. documented that perceived quality of work life was significantly greater for people practicing relaxation techniques, thus reducing stress felt at work (84). Also of interest, M. M. Delmonte found in one of his research done in 1985, that people who practice meditation tend to have a healthier psychological profile than those who do not (85). It would also reasonable to say that a person practicing such techniques may be more sensitive to his or her health, thus less susceptible to consume psychotropic medications as a way to relieve or compensate for job strain.

A good, fair or poor perception of health was also very strongly associated with psychotropic drug use compared with those who perceived it as excellent or very good. Although this factor is broad and general by definition, it is a very good indicator of health in the population. It reflects the global appreciation of one’s health, integrating self-knowledge and experiences with health and sickness. This indicator is reliable and accurate because of its correlation with other factors representing specific health dimensions as determined by many studies (78, 86-89).

In order to assess job strain, many researchers used the Karasek’s demands-control model (19, 23-26). All studies reviewed but one (24) found an association between job strain and psychotropic drug consumption. The association was also observed in most studies using other types of measures of job stressors such as bad relations with colleagues or superiors (17, 21, 24, 57), physical demands at work (22, 25, 57), lack of employment security (25), employment status (18, 19), occupation (19, 53), change in working conditions (66), seniority (18), salary or family income (19, 46, 50, 53, 54), shift work (21, 57, 90) or the number of days of absence from work during the last year (49, 51). Using a part of the same cohort as for the present study, Moisan et al. made a cross-sectional analysis with the data obtained at the beginning of follow-up from 1991 to 1993. Their results supported the hypothesis that exposure to high job strain significantly increases the odds of psychotropic drug consumption (19). Social support at work did not modify the association. Age, education level, stressful events that occurred during the last 12 months and level of physical activity were also positively associated with psychotropic drug consumption (19).

Having followed a group of workers from 1991 to 2001 and used only persons who stayed within the cohort during follow-up, cross-sectional analysis could have amplified the healthy worker effect. There is a good chance that the sickest persons in our population sample were lost at follow up by quitting or changing work. In such a case, they were not part of the eligible population for the current analysis. This would leave us with a healthier than normal worker group. This selection bias is likely to be non-differential for almost all the cofactors measured, but could be differential for some like age, because illnesses are more likely to occur with advancing age.

Since we had to exclude individuals who did not satisfy the selection criteria (i.e.: was a white-collar worker, worked full time, held their present position for more than a year) there might be a selection bias because in the follow-up analysis, the study sample was composed of “survivors” only. Ideally, our population should have been the same for baseline and follow-up analyses. As explained in the preceding paragraph, some participants had left during the 10 years between the two analyses. Theoretically, participants’ withdrawal should be equal in all subgroups of our study population but because of the possible selection bias discussed above, it may have caused an underestimation of the true odds ratio observed in the follow-up analysis. This could be an explanation for the discordant results showing a significant association between job strain and psychotropic drug use at baseline analysis but any, about 10 years later, at follow-up analysis.

In our study, job stressors were measured using the 18 item questionnaire recommended by Karasek. Even if the validity of this questionnaire has been well established, participants might have been misclassified due to the cut-off point we referred to (sample of workers in the province of Quebec). However, the most likely bias is a non-differential classification bias, which would most likely result in the association measured to be underestimated (91).

Siegrist’s extrinsic efforts-rewards imbalance and rewards alone model gave significant results in regard to psychotropic drug consumption. Since extrinsic efforts were measured using a proxy measure composed of the elements contained in Karasek’s psychological demands, it is plausible that the rewards measurement could be more sensitive in discriminating factors that could influence psychotropic drug consumption than decision latitude. Compared to decision latitude, the rewards scale includes measurements of expectations a worker could have with regard to his or her job. This aspect, not covered by decision latitude, seems to be of importance, since results differed noticeably between Karasek’s and Siegrist’s models. Even if further comparison of these models is essential, results presented here suggest that rewards and expectations from work play a non-negligible role in regard to psychotropic drug consumption.

Also of interest, intrinsic efforts (overcommitment) have not significantly influenced the relation between an imbalanced extrinsic efforts-rewards ratio and the consumption of psychotropic drugs (no modifying effects). This would suggest that psychotropic drug consumption is less influenced by personality traits than it is with environmental consideration. This finding, that will need to be corroborated, is of importance because an intervention to prevent psychotropic drug consumption would have to be principally oriented on the work environment to be optimally effective.

The only analyses showing a strong moderating effect of gender were the one done with antidepressants (Karasek’s and Siegrist’s models). Even if our general findings for psychotropic drugs and benzodiazepines use are not consistent with the literature, they seem to indicate a stronger tendency for women, compared to men, to consume antidepressants when facing job strain or an extrinsic efforts-rewards imbalance.

A certain number of limitations need to be acknowledged. The eligible population was taken from a cohort study on cardiovascular diseases. It is therefore possible that our study attracted more people with such diseases. The sensitivity to job stress of this subgroup of people may be higher than the general working population resulting most likely in an overestimation of the association between job strain and psychotropic drug use. Since we observed no significant relationship between those last two factors, a selection bias due to the origins of our study population is unlikely to have occurred. External validity of our analysis may also be limited since it is possible that exposure to job strain and stress felt at work may not be representative of what’s happening in the general working force of the entire province of Quebec or elsewhere.

A recall bias for psychotropic drug use is unlikely, given that the participants were questioned on the use of such drugs during the previous 48 hours only. The usage of these kinds of drugs being socially discouraged, an underestimation of the real consumption pattern may have occurred, underrating the real prevalence. This phenomenon affects all studies of this kind and can cause the association with job strain to be underestimated. Using a 48 hours recall period may also cause an underestimation of the true prevalence of psychotropic drug consumption because stress due to job strain can, in some cases, be temporary. For example, using a 48 hours recall period for somebody who suffered from exposure to job strain for a short period a couple of months ago would result in a negative answer even if the participant used psychotropic medications at that time. It is a limitation related to the transversal design of this study.

A confounding bias may be present since there was no adjustment made for health problems because this variable is known to be associated with psychotropic drug use. Mental health problems such as depression or anxiety as well as physical health problems such as a history of myocardial infarction may be in the causal chain of events between exposure to job strain and psychotropic drug use. In such a situation, it is prudent not to include this variable when making adjustments.

The deleterious confusing effect of a high social support at and outside of work observed in our study is contrary to results from most references found in the literature. This inverse association could partly be explained by the questions used to assess social support at work in our study. In the self-administered questionnaire, questions about social support at work were mostly about the quantity and not the quality of social relations. A person could have a lot of friends or contacts but may not get real help or support from them. In that situation, reports of high social support at work do not mean that a person will be able to use his social network to reduce the stress felt at work. Perhaps a different formulation of the questions used in our study could have prevented the effect observed here. The questions used to assess social support outside of work measured both the quantity and certain qualitative aspects of relationships. Even with this dimension added, we observed a significant deleterious effect with a high social support outside of work. This could indicate that our questions may not be pertinent or precise enough to describe social support outside work. Another possible explanation could be that people with an efficacious social network would probably tend to keep working in a greater proportion than people who do not. In such a case, these persons would tend to be overrepresented in our study population thus giving the kind of relation observed here. The relation observed between social support at work and outside of work could also simply mean that our population is not representative of a general workers population and thus gave distorted results.

Results found using Siegrist’s extrinsic efforts - rewards model are very interesting and unveil new and also solid evidence that environment plays a primordial role in the genesis of psychotropic drug use in response to work conditions. Since we are the first research team to use this theoretical context with regards to use of psychotropic drugs, we hope that further studies using a comparable context will follow.

Even if unanimity between studies and scientific communities is not yet achieved, it is imperative to continue research about this phenomenon and particularly with longitudinal designs. Health problems and work absenteeism (16, 59) largely justify the continuation of efforts about understanding the factors determining psychotropic drug use.

Figure 1 Selection criteria

Results tables - Karasek’s job strain model

* Multivariate analysis are adjusted for social support at work, social support off work, age, gender, education, family income, employment status, occupation, stress associated with events that occurred during the last 12 months, practice of relaxation/meditation technique, perception of health, cigarette smoking, alcohol consumption and physical activity.

§ Reference category

† PD = Psychological demands, DL = Decision latitude, + = High exposition, - = Low exposition.

* Multivariate analyses are adjusted for social support at work, social support off work, age, gender, education, family income, employment status, occupation, stress associated with events that occurred during the last 12 months, practice of relaxation/meditation technique, perception of health, cigarette smoking, alcohol consumption and physical activity.

§ Reference category

† PD = Psychological demands, DL = Decision latitude, + = High exposition, - = Low exposition.

* Multivariate analysis are adjusted for social support at work, social support off work, age, gender, education, family income, employment status, occupation, stress associated with events that occurred during the last 12 months, practice of relaxation/meditation technique, perception of health, cigarette smoking, alcohol consumption and physical activity.

§ Reference category

† PD = Psychological demands, DL = Decision latitude, + = High exposition, - = Low exposition.

* Multivariate analyses are adjusted for social support at work, social support off work, age, gender, education, family income, employment status, occupation, stress associated with events that occurred during the last 12 months, practice of relaxation/meditation technique, perception of health, cigarette smoking, alcohol consumption and physical activity.

§ Reference category

Results tables - Siegrist’s extrinsic efforts - rewards model

* Multivariate analyses are adjusted for social support at work, social support off work, age, gender, education, family income, employment status, occupation, stress associated with events that occurred during the last 12 months, practice of relaxation/meditation technique, perception of health, cigarette smoking, alcohol consumption and physical activity.

§ Reference category

* Multivariate analyses are adjusted for social support at work, social support off work, age, gender, education, family income, employment status, occupation, stress associated with events that occurred during the last 12 months, practice of relaxation/meditation technique, perception of health, cigarette smoking, alcohol consumption and physical activity.

§ Reference category

* Multivariate analyses are adjusted for social support at work, social support off work, age, gender, education, family income, employment status, occupation, stress associated with events that occurred during the last 12 months, practice of relaxation/meditation technique, perception of health, cigarette smoking, alcohol consumption and physical activity.

§ Reference category

* Multivariate analyses are adjusted for social support at work, social support off work, age, gender, education, family income, employment status, occupation, stress associated with events that occurred during the last 12 months, practice of relaxation/meditation technique, perception of health, cigarette smoking, alcohol consumption and physical activity.

§ Reference category

* Multivariate analyses are adjusted for social support at work, social support off work, age, gender, education, family income, employment status, occupation, stress associated with events that occurred during the last 12 months, practice of relaxation/meditation technique, perception of health, cigarette smoking, alcohol consumption and physical activity.

§ Reference category

* Multivariate analyses are adjusted for social support at work, social support off work, age, gender, education, family income, employment status, occupation, stress associated with events that occurred during the last 12 months, practice of relaxation/meditation technique, perception of health, cigarette smoking, alcohol consumption and physical activity.

§ Reference category

* Multivariate analyses are adjusted for social support at work, social support off work, age, gender, education, family income, employment status, occupation, stress associated with events that occurred during the last 12 months, practice of relaxation/meditation technique, perception of health, cigarette smoking, alcohol consumption and physical activity.

§ Reference category

* Multivariate analyses are adjusted for social support at work, social support off work, age, gender, education, family income, employment status, occupation, stress associated with events that occurred during the last 12 months, practice of relaxation/meditation technique, perception of health, cigarette smoking, alcohol consumption and physical activity.

§ Reference category

* Multivariate analyses are adjusted for social support at work, social support off work, age, gender, education, family income, employment status, occupation, stress associated with events that occurred during the last 12 months, practice of relaxation/meditation technique, perception of health, cigarette smoking, alcohol consumption and physical activity.

§ Reference category