Pharmacological Neuroenhancement Among German University Students: Identication of Potential Risk Groups and its Relation to Psychological, Psychosocial, and Health Behavioral Factors

Aiming to develop and implement intervention strategies targeting pharmacological neuroenhancement (PN) among university students more specically, we i) assessed the prevalence of PN among German university students, ii) identied potential sociodemographic and study-related risk groups, and iii), investigated sociodemographic, psychological, study-related psychosocial, general psychosocial and health behavior related factors predicting the 12-month prevalence of PN. Therefore, a cross-sectional online survey was administered to all students of the University of Mainz, Germany. A binary logistic regression with stepwise inclusion of the ve variable groups was performed to predict PN. A total of N = 4,351 students participated in the survey of which N = 3,984 answered the question with regard to PN. Of these, 10.4% had used one substance for PN at least once in the past 12 months. The regression revealed 13 variables that were signicantly related to the 12-month prevalence of PN. Specically, the group of health behavior variables had the strongest inuence on the explained variance of PN. Therefore, an approach to the prevention of PN should be multifactorial so that it addresses social conditions, as well as education on substance use and healthy behaviors in terms of non-pharmacological strategies as alternatives of PN.


Introduction
The term "pharmacological neuroenhancement" (PN), also called "pharmacological cognitive enhancement", is generally de ned as the use of illicit or prescription drugs by healthy individuals for cognitive-enhancing purposes 1-3 such as enhancing alertness, attention, concentration, memory and also mood 4,5 . According to this de nition, the so called soft neuroenhancers (e.g. energy drinks, caffeine tablets, ginkgo biloba) are not included. There are many inconsistencies and differences in the de nition of this issue 6,7 but a full discussion of these would go beyond the scope of this research.
In the past decade, a considerable number of studies demonstrated that PN was not uncommon in western populations. For example, a study among 102,000 adults in the United States (US) reported a 12month prevalence of 2.1% for the use of prescription stimulants in the general population, and of 6.4% among participants in the age between 18 and 29 years. Mostly reported reasons (78.2%) for the use were improving alertness, concentration or help to study 8 . Furthermore, the global drug survey from 2017, a large cross-sectional study performed in 15 countries including 29,758 participants, reported a 12month prevalence of 6.6% for the use of prescription stimulants for cognitive enhancement, with large differences in the prevalence rates between countries 9 . Another study from Europe reported a lifetime prevalence for PN of 4.0% among 10,171 participants from the general population of Switzerland 10 .
Furthermore, other studies assessed the prevalence of PN in speci c populations and occupations, such as Franke et al. 5 or Dietz et al. 4 , reporting a lifetime prevalence for PN of 8.9% among German-speaking surgeons, and of 19% among German-speaking economists.
A very well examined group with an increased risk for PN is the collective of university students. For example, a large study comparing the non-medical use of prescription stimulants between US college students and respondents of the same age not enrolled in college (N = 15,454), showed that college students used prescription stimulants more often (OR 1.28, 95% CI 1.05, 1.56) compared to non-students of the same age group 11 . Moreover, within a comprehensive review and meta-analysis, Benson et al. 12 reported 12-month prevalences for the use of prescription stimulants between 5% and 35% among college students in the US, demonstrating large heterogeneity in the range of these prevalence rates. Studies among university students in Western Europe obtained results in a similar range. For example, lifetime prevalences for PN of 7.8% among 6,275 Swiss students 13 , 3.2% among Norwegian students 14 , and 19.2% in a sample of students from the United Kingdom 15 were reported. The same tendencies appear among German university students, as lifetime prevalences of 4.6% 16 , and 12-month prevalence estimates between 11.9% 17 and 20% 18 were reported.
From a public health point of view, the above-mentioned gures for the use of PN, especially in university students, are alarming, because PN appears to be associated with physiological and psychological side effects, may increase mortality, and can lead to addiction [19][20][21][22][23][24] . Understanding the conditions and factors predicting PN, especially among the severely affected collective of university students, contributes to evidence-based planning of PN-prevention strategies, because effective programs have to target factors related to PN. Therefore, potential correlates (factors that are associated) or determinants (factors with a causal relationship) of PN necessarily have to be investigated 25 . In this context, some research already investigated potential variables being related to PN such as sociodemographic aspects 17,18 , psychological factors such as stress 26-29 , as well as different demands and resources [30][31][32][33] . Furthermore, study-related psychosocial factors such as perceived academic bene ts [34][35][36][37][38] , general psychosocial factors 39 , as well as health behavior 17,40 have also been associated with PN.
Among this body of research, studies in university students examining the role of explanatory variables in an adequate sample size are rare. Furthermore, to the best of our knowledge we are not aware of any study investigating the relation between PN and sociodemographic factors, psychological factors, studyrelated and general psychosocial factors, as well as health behavior related factors in one model. In addition, Faraone et al. 7 summed up that, among other aspects, because of limited data availability, and variations in describing the use of PN, more research would be needed to identify potential risk groups of PN use, and to develop effective prevention and treatment interventions.
To conclude, empirical studies addressing PN among university students are heterogenous regarding their methodology and results [13][14][15][16][17][18]41 . Moreover, there is a considerable lack of knowledge with regard to potential factors that might predict PN as well as to the identi cation of potential study-related risk groups. Therefore, within the present study, we addressed these issues and i) assessed the prevalence of PN among German university students aiming to ii) identify potential sociodemographic and studyrelated risk groups, especially with regard to age, gender, eld of study, semester, aspired degree, and iii) investigate factors related to PN by putting sociodemographic factors, psychological factors, study-related and general psychosocial factors, as well as health behavior related factors in one stepwise regression model. This may enable us to identify more general factors as well as speci c variables that might be more or less strongly related to PN. These results may be used to iv) develop and implement intervention strategies targeting PN among university students more speci cally.

Methods
Study design and survey procedure A cross-sectional online health survey was administered to all students of the University of Mainz, Germany (31,213) in summer term (June and July) 2019 as part of an ongoing project on health promotion among students ("Healthy Campus Mainz"). Students were invited to participate through the university's central mailing list. Reminder emails were sent four times. In an introduction at the beginning of the online questionnaire, the background and purpose of the study were shortly explained, followed by a statement that participation would be anonymous and voluntary. Informed consent was obtained written at the beginning of the survey. A total number of N = 4,351 students participated in the survey, demonstrating a response rate of 13.9% of the university's total student population at that time. Approval

Measures
The online survey covered a wide range of health-related topics containing approximately 270 items.
Established and validated instruments were used whenever feasible and self-developed scales were used as little as possible. A whole list of all surveyed topics and items is given by Reichel et al. 42 . To predict PN for the present paper, 55 independent variables with regard to the research questions were selected and classi ed in 5 different groups (according to the factor groups of current research, as described in the introduction): sociodemographic variables (14 variables, e.g. gender, age, semester, eld of study), psychological variables (6 variables, e.g. depressive symptoms, emotional exhaustion), study-related psychosocial variables (17 variables, e.g. social support by fellow students, self-e cacy expectance), general psychosocial variables (5 variables e.g. self-criticism, impulsiveness), and health behavior related variables (13 variables e.g. alcohol use, healthy diet, physical activity). A list of the speci c variables, scales, and items used for the present paper as well as the respective references is given in table 1.
The prevalence of PN was assessed according to Dietz et al. 4 . The translated question regarding PN was: "Have you ever used the following substance/-s without medical necessity, for the purpose of enhancing your cognitive performance or to better handle your studies (not for reasons of enjoyment)?". The following illicit or prescription drugs could be selected via multiple choice, and each with the scale 'never', 'within the last 30 days', 'within the last 12 months', or 'more than 12 months ago': methylphenidate (e.g. Ritalin®), amphetamine preparation (e.g. Adderall®), atomoxetine (e.g. Strattera®), moda nil (e.g. Provigil®), ecstasy, ephedrine, cocaine, illicit amphetamines (e.g. Speed), crystal meth, cannabis, and 'other substances'. Table 1 List of independent variables differentiated in the ve category groups. 'Count' means the count of variables if 2 or more variables of the same scale are listed together.

Results
A total of N = 3,984 participants answered the question with regard to PN and were included in the analyses. Mean age of the sample was 23.8 years (SD = 4.3 years) and 71.3% (n = 2,842) of the participants were female. With regard to study-related characteristics, 16.3% (n = 650) of the participants were rst-year students, 52.4% (n = 2086) were studying for a bachelor's degree, 21.2% (n = 844) for a master's degree, 22.0% (n = 876) were aiming for a German state examination (e.g., law and medical students and students of teaching professions), and 3.5% (n = 139) were PhD students. All sociodemographic and study-related characteristics of the participants are presented in table 2. Table 2 Characteristics of the study sample. SD: standard deviation; STEM: science, technology, engineering, and mathematics.   The 12-month prevalence of PN varied across sociodemographic and study-related groups. In male students, the prevalence was signi cantly higher (13.2%) compared to female students (9.3%, p < .001).
First-year students showed a slightly higher 12-month prevalence (12.2%) compared to higher semester students (9.9%), and a signi cant difference was identi ed between bachelor (12.8%), and master (7.7%) students (p = .010 and p = .025, respectively). The 12-month prevalences for all sociodemographic and study-related variables are given in  The relation between the 12-month prevalence of PN and sociodemographic, psychological, study-related psychosocial, general psychosocial, and health-behavioral factors The sample size for pre-testing and binary logistic regression after the above-described protocol of dichotomization was n = 3,800. Pre-testing (supplemental tables 1 and 2) revealed 29 variables that were signi cantly associated with the 12-month prevalence of PN. Due to the number of variables, case processing of the logistic regression included n = 3,584 cases into the analysis, demonstrating an appropriate size respectively events per variable. However, the rst run of the regression analysis demonstrated that the variable "degree" had to be removed from the model, because it showed p-values between .999 and 1 in all categories.
Subsequently, the overall model of the stepwise binary logistic regression was statistically signi cant, χ²  Male students had a higher likelihood of PN within the last 12 months, compared to female students.
'Depressive symptoms' was positively related to the 12-month prevalence of PN. Likewise, the more the respondents regularly stayed away from their study events (i.e. absenteeism) and the more excessive they rated their demands, the higher was the likelihood of PN within the last 12 months. Negatively related variables were 'social support by fellow students' and 'self-criticism': the greater their expression, the less likely was PN within the last 12 months. Among health behavior variables, sticking to a healthy diet, extent of moderate-vigorous physical activity, risky alcohol use, currently smoking cigarettes, consumption of coffee, caffeine tablets, coke, and ginkgo biloba were positively related to the 12-month prevalence of PN. Table 5 Signi cant predictors of the 12-month prevalence of pharmacological neuroenhancement in a binary logistic regression analysis with stepwise inclusion of the 5 independent variable groups: sociodemographic, psychological, study-related psychosocial, general psychosocial and health behavior related variables.

Discussion
Using a cross-sectional approach, the present study aimed to i) assess the prevalence of PN among university students, ii) identify potential risk groups for PN and iii) investigate the explanatory role of sociodemographic, psychological, study-related and general psychosocial, and health-behavioral variables on PN among a large sample of German university students, in order to iv) contribute to the development and implementation of intervention strategies targeting PN among university students more speci cally.
The overall 12-month prevalence for PN was 10.4%. This prevalence is approximately in the middle of the reported prevalences for university students from western European countries [13][14][15][16][17][18] . As stated above, these differences among reported prevalences may be caused by various methodological aspects (e.g. de nition of PN, survey technique or period of reported prevalence).
With regard to the second research question, namely the identi cation of potential risk groups for PN in the collective of university students, male students showed a signi cantly higher risk for PN compared to female students. This nding is in line with previous studies 12,69,70 , showing that substance use appears to be more common in males than in females in different populations. With regard to study-related risk groups, rst-year students and bachelor students were of increased risk for PN. This implies that PN is practiced early during studies, con rming the ndings of Dietz et al. 18 . Furthermore, the prevalence of PN varied between different elds of studies: especially students of 'social sciences, media and sport' had a higher risk for PN compared to students from other elds of study. A possible explanation may be that the use of nutritional supplements (e.g., vitamins, minerals, herbals, caffeine or creatine) is common in the eld of sports and discussed to provide a gateway to the use of illicit drugs [71][72][73][74] .
In view of the third research question, namely the investigation of the explanatory role of sociodemographic, psychological, study-related and general psychosocial, and health-behavioral variables, gender was the only signi cant sociodemographic variable in the regression model. This is in accordance with results of the contingency-analysis (second research question), that male students had a higher likelihood of PN within the last 12 months compared to female students. Among the group of psychological variables 'depressive symptoms' showed a small positive association with PN.
Surprisingly, other psychological variables such as 'general anxiety', 'social anxiety' or 'loneliness' were not signi cantly related to PN in our model, although previous studies reported associations 69,[75][76][77] . In contrast to our study, these studies investigated the association of the different psychological variables and PN more isolated and not in a large model, like we did in the present study. Concerning the studyrelated and general psychosocial variables, a small protective effect of 'social support by fellow students' and 'self-criticism' for the use of PN was revealed. To the best of our knowledge, these variables were not investigated before when predicting PN. In this context, more 'social support by fellow students' also could be associated with more organized studying or less competition and therefore reduce stress and the subjective need of PN to increase academic performance 27 . The preventive effect of 'self-criticism' could be due to higher personal standards and perfectionism of more self-critical individuals 78 and therefore, they might rather try to reach their academic performance goals by themselves and without external help through PN. Likewise, a potential explanation for 'absenteeism' predicting PN is that staying away from lectures increases the pressure to catch up on learning material and to successfully pass an exam, so that PN may be used to cope with these demands. In the context of these results, it is also plausible that PN is a coping strategy in a vicious cycle of depression, missing in lectures, low social support and maybe other forms of substance use or self-endangering behaviors. In general, selfendangering behaviors (e.g. presenteeism or prolonging working hours) represent maladaptive coping strategies and have previously been associated with higher quantitative demands and autonomy (in a ushaped connection) among students 79 .
Interestingly, the group of health behavior related variables (healthy diet, extent of moderate-vigorous physical activity, alcohol use, smoking cigarettes and using soft neuroenhancing substances) contributed the most to the explanation of PN in our model. Surprisingly, although 'healthy diet' appears as a healthprotective factor for several issues 80 , the present results indicate that 'healthy diet' (for de nition see table 1) increases the likelihood of PN. A possible explanation could be that healthy diet is also associated with cognitive bene ts 81 , and consequently it could be used as a co-strategy for neuroenhancement. But sticking to a healthy diet might also be biased by subjective assessment of one's own diet and therefore may be an indicator of restrictive diets or eating disorders that have been investigated to positively relate to PN 40,82 . The extent of moderate-vigorous physical activity showed a very small positive association with PN. This might be linked with the higher prevalence of PN in the cohort of students from the faculty of 'social sciences, media and sport'. Furthermore, PN was strongly associated with other forms of drug involvement (alcohol use, smoking cigarettes). Whereas most of the nal model's odds ratios were small, for the consumption of soft neuroenhancers (coffee, caffeine tablets, coke, and ginkgo biloba), we obtained medium effects on the likelihood of PN within the past 12 months. While other forms of drug use, like drinking and smoking, may be a reason for decreased academic performance 83 and reinforce the apparent necessity of PN, these soft neuroenhancers may provide a gateway to PN 1,74,84 . Especially for the use of caffeine tablets, relations to PN were already stated by other studies 17,85 . But besides these potential dangers of shifting from legal to illicit substance use, the consumption of high dosages of caffeine is also associated with adverse health-effects 86-89 . However, it should be noted that this block of health behavior related variables includes other forms of substance use and since PN also represents a form of substance use, a greater explanation of variance by this step seems plausible.
Additionally, it is worth noting that our model did not support a signi cant in uence of the amount of social media use on the past 12-month prevalence of PN, referring to previous research that recommended investigating this relationship 6 .
Since university students represent the executives, decision makers and also parents of tomorrow, health promotion in university students may not only be bene cial for the target group, but also bene t the general society 90 . In order to iv) develop health promotion and prevention programs of high quality, such programs should be evidence based. Since we revealed a higher prevalence of PN in rst-year and bachelor students, prevention of PN should start early during studies or even at the end of school.
Therefore, more research on the prevalence of PN among pupils, especially in graduation classes, would also be bene cial. Consequently, given that the present results show that PN is predicted particularly by the use of soft neuroenhancers, strategies tailored to educate on the use and effects of these substances may also help to prevent the more harmful use of drugs for the same purpose. Prevention strategies on general consumption of intoxicants, like drinking and smoking, may also decrease the risk of engaging in PN, since this study demonstrated the contributing effect of a risky alcohol consumption, or smoking cigarettes. Moreover, since some individuals seem to use PN without critical re ection of potential consequences 38 , students should be educated about the limited e cacy of PN in healthy individuals 91,92 and that PN is not associated with better marks or increased academic performance 93 . Additionally, more research should focus on the role of social support in the context of PN, because cultivating and developing social support such as networks of communication and mutual obligation acts as a great resource [94][95][96] , not only with respect to the prevention of PN. Interventions that are targeted at the risk groups and use a multifactorial approach could lead to effective prevention of PN in future. Such a multifactorial prevention approach should therefore address social conditions, and educate on substance use as well as on healthy behaviors to increase cognitive performance -such as nutrition, physical activity and mindfulness 81,97-99 -and how to adopt these behaviors as habits.
With regard to potential limitations of the present study, one could argue that whenever sensitive topics are studied, participants often react in a way that negatively affects the validity of study results (underreporting and non-responding) due to hesitating to provide compromising information about themselves 100,101 . Therefore, other studies used indirect survey techniques such as the randomized response technique (RRT) for the assessment of socially desirable questions such as PN 102,103 . Since the 12-month prevalence of PN of 10.4% in our survey is comparable to those of RRT-surveys 17 , it may imply that an online poll is subjectively perceived as anonymous and private and therefore provides comparable results to RRTs. In general, our online survey aimed to reach all students of Mainz University.
Nevertheless, as participation was voluntary, we cannot exclude a certain selection bias in our sample. For example, health interested students and students of health-related disciplines might be more likely to participate in a health survey. Because of this potential bias, the results regarding the prevalence of PN have to be interpreted with caution. Nevertheless, a strength lies in the robust associations on individual level. It also has to be noted that our study had a cross-sectional design, and therefore, no causality of the analyzed conditions can be stated.

Conclusion
This study reveals that the 12-month prevalence of PN among German university students differs in regard to sociodemographic and study-related groups, with speci c risk groups being males, rst study year students, and the study elds of 'social sciences, media or sport', and 'law or economics'. Therefore, future studies should be performed with respect to the prevalence of PN in school graduation classes and prevention of PN should start early during studies or even at the end of school. This study further reveals that a model with groups of psychological, psychosocial and health behavior related variables is suitable to explain the 12-month prevalence of PN. In that model, speci cally the group of health behavior variables has the strongest in uence on the explained variance of PN. Therefore, an approach to the prevention of PN should be multifactorial so that it addresses social conditions, as well as education on substance use and healthy behaviors to increase cognitive performance and cope with stress. Students should be aware of and be able to habitually implement non-pharmacological coping-strategies 104 that can help to increase cognitive performance and mood such as physical activity 105 , nutrition, and relaxing or mindfulness techniques.

Data Availability
The datasets generated and analyzed during the current study are stored on the server of the University Medical Center of the University of Mainz (European server) and are available from the corresponding author on reasonable request. The manuscript has been read and approved by all named authors.