Risk behaviour among women using combined oral contraception

Dnr 2013/161 Risk behaviour among women using combined oral contraception Sara Ilic Supervisor: Inger Sundström Poromaa Department of Women s and Children s Health, Reproductive Health Uppsala University 2018-09-16

Abbreviations... 3 Abstract... 4 Populärvetenskaplig sammanfattning... 5 1 Introduction... 6 1.1 Risk behaviour... 6 1.2 Decision making... 6 1.3 Impulsivity and development of the frontal lobes... 7 1.4 Iowa Gambling Task... 7 1.5 Combined oral contraception... 8 1.6 Research Question... 10 2 Material and method... 11 2.1 Participants... 11 2.2 Study design... 14 2.3 Iowa Gambling Task... 14 2.4 Statistical analyses... 15 3 Results... 17 3.1 Subjects... 17 3.2 Iowa Gambling Task... 19 3.3 Outcome depending on treatment... 20 4 Discussion... 24 4.1 Introduction, method and results... 24 4.2 Benefits and limitations... 28 4.3 Future aspects... 29 4.4 Conclusion... 29 5 Acknowledgements... 30 6 References... 31 2

Abbreviations ACC Anterior cingulate cortex AUDIT Alcohol use disorders identification test BMI Body mass index COC Combined oral contraceptive DRSP Daily record of severity of problems E2 Estradiol 5-HT Serotonin IGT Iowa gambling task IQ Intelligence quotient IQR Interquartile range MADRS-S Self-rated version of Montgomery-Åsberg depression rating scale NOMAC Nomegestrolacetate OFC Orbitofrontal cortex PFC Prefrontal cortex PMDD Premenstrual dysphoric disorder PMS Premenstrual syndrome RCT Randomised controlled trial SD Standard deviation SSP Swedish universities scale of personalities SSRI Selective serotonin reuptake inhibitor 3

Abstract Objective: There are studies about impulsiveness among women during the menstrual cycle, but little is known about how hormonal contraceptives affect risk taking. Recent research has demonstrated improvements in judgement and concentration among combined oral contraceptive (COC) users. Given the lack of randomised controlled trials in this area, we aimed to investigate whether women on COC display decreased risk behaviour compared to women taking placebo. Methods: The study is an investigator-initiated, double-blinded, placebo-controlled randomised trial. Women were randomised to take COC or placebo during three treatment cycles. At the last treatment visit, the participants in Uppsala were asked to do the Iowa Gambling Task (IGT), a test to measure risky decision making. Results: The IGT results from 40 women were analysed. Overall, the results among the groups were similar. However, only the women taking COC made a shift in preference for advantageous decks (F (1, 21) = 5.28; p = 0.032). Also, in block 2 the women on COC selected more advantageous decks compared to the women on placebo (p = 0.041). Conclusion: Overall, the groups results were alike. Yet, only the women on COC shifted to a strategic approach. This finding indicates that these women understood the task faster. 4

Populärvetenskaplig sammanfattning Impulsivitet kopplat till menscykeln är ett populärt forskningsområde. Studier visar förbättrat omdöme och koncentrationsförmåga bland p-pilleranvändare men det saknas randomiserade kontrollerade studier inom området. Den här delstudien undersöker huruvida riskbeteende påverkas av kombinerade p-piller. Riskbeteende undersöktes med hjälp av Iowa Gambling Task (IGT), ett test för att undersöka riskbeteende och beslutsförmåga. Kvinnorna randomiserades till att ta p-pillret Zoely eller placebo under tre behandlingscykler. Vid slutbesöket tillfrågades kvinnorna i Uppsala om de ville delta i IGT-studien. Deltagarna placerades framför en datorskärm där fyra olika kortlekar visades. För varje valt kort kunde de vinna eller förlorade i fiktiva i-dollar. Vissa kortlekar var mer fördelaktiga än andra i längden. Om deltagarna valde kort från endast de fördelaktiga högarna resulterade det i långsiktiga vinster, medan upprepade kortval från de ofördelaktiga högarna medförde samlade förluster. Innan spelet startade erhölls instruktionen att försöka spela för att vinna så mycket i-dollar som möjligt. Totalt var det 40 friska kvinnor i åldern 18 35 år som deltog i studien, tjugotvå av dessa var randomiserades till p-piller medan de övriga 18 fick placebo. Grupperna var jämförbara och de statistiska analyserna visade att majoriteten av IGT-resultaten var likvärdiga i de två grupperna. Det man däremot såg var att kvinnorna som fick Zoely snabbare bytte taktik och började välja fler kort från de fördelaktiga korthögarna. Detta indikerar att de möjligtvis förstod testet snabbare. Då så pass få deltagare deltog i studien och resultaten inte var entydiga kan vi inte säkert uttala oss huruvida kvinnors riskbeteende påverkas av p-piller. Framtida forskning kräver fler deltagare och mätinstrument. Nya studier skulle antingen kunna förkasta de rön som råder gällande minskad impulsivitet bland p-pilleranvändare, eller ge resultat som medför nya indikationer för förskrivning av p-piller. 5

1 Introduction 1.1 Risk behaviour According to the Swedish encyclopaedia, risk behaviour is defined as a lifestyle that entails risk and above all health risks (1). Nicotine use, alcohol consumption, unhealthy gambling and sex without protection are some examples. A lot of research interest has been devoted to explore how a subject s risk behaviour affects the use of contraception. Some examples are the studies by Leon- Larios and Maciás-Seda (2), Papas (3) and Saffier (4). The study by Leon-Larios and Macías-Seda examines sexual behaviour among university students in Seville and suggests improved sexual behaviour with age, female gender, education of sexuality and contraception and education in health sciences (2). As regards the study by Saffier, it reviews published research about the risk of getting the HIV virus among adolescents in Brazil (4). Some of the identified risk factors according to the study are age, the usage of drugs and alcohol, socioeconomic status, being married, geography, sexual and reproductive history, infection history and mental health (4). These risk factors are consistent with suggested risk factors that contribute to increased risk behaviour in other studies; low socioeconomic status (5) (6), substance use (5), low educational level, poor social support, comorbid mood conditions, childhood difficulties and psychiatric diagnoses (6). Further, risk behaviour affects decision making. In a review, the effect of emotional, social and environmental influences on decision making are underlined, especially in situations that entails risk or uncertainty (7). Clearly, using hormonal contraception is a useful way of reducing the risk of unplanned pregnancies, and may, in itself also be a sign of low risk behaviour. However, it is not known if hormonal contraceptives, per se, influence the risk taking of the women who are using them. When searching in PubMed for articles on how the use of combined oral contraceptive affects risk behaviour, the lack of research is evident and no randomised controlled trial (RCT) was found. 1.2 Decision making Executive abilities in terms of judgement, set shifting, planning, anticipation, reasoning and, least but not last, decision making, are important to manage assignments that test cognition or behaviour (8). Decision making is a complex matter that requires much of the person concerned. To make a decision, the individual must manage a great amount of inflow of information and be flexible in planning as the terms and conditions constantly change. Furthermore, decision making includes consideration of consequences in the short- and long term (9). When people make choices that affect future possibilities, short-time rewards are often preferred over long-time rewards (10). 6

Accordingly, the application of behaviour control is necessary to optimize long-term profits (9). Regardless, there is always an uncertainty of the outcome and even a well-made decision can result in unexpected consequences (11). Neuroimaging findings demonstrate that several brain regions are associated with different subtypes of decision making. Reward-based choices, for example, are associated with the limbic system, probabilistic-based decisions are linked to activity in the parietal cortex and decision making comprising risk or uncertainty is linked to the medial prefrontal cortex (PFC), orbitofrontal cortex (OFC), rostral anterior cingulate cortex (ACC) and caudate (12). 1.3 Impulsivity and development of the frontal lobes Impulsivity comprise poor inhibitory control, attention difficulties, an incapacity to validate consequences and/or an inability to avoid small short-term rewards in favour for superior long-term rewards. These issues are more often seen in adolescents than in adults. Moreover, the adolescence is a period when important changes occur in the brain in order to adapt to adequate behaviour fit for the milieu and to promote the protection of the group or family. Some examples of neurochemical adaptions that occur during the adolescence are elevated dopamine and serotonin (5-HT) levels and inputs to the PFC, increased cholinergic innervations of the PFC and a changed reaction to cortisol (13). In contrast to impulsivity, decision making entails attention. However, with repetition, less attention is needed and more quick decisions are made. As for cognitive assignments, such as attentiveness, memory and the choice to respond to a present stimulus, the frontal lobes are important. Experiences from the environment and practice, such as studying for an examination, affects the cerebral cortex and change the structure in the brain (13). Additionally, education is suggested to be associated with higher intelligence quotient (IQ) scores both earlier and later in life (14). All in all, the brain is not considered to be fully developed until the age of 25 (15). The frontal lobes, which are important for executive functions, mature during late adolescence (13). 1.4 Iowa Gambling Task The Iowa Gambling Task (IGT) is a frequently used test to measure decision-making (16) under risk (17). When performing the task, the subject is placed in front of a computer screen where four different decks of cards are displayed. When picking a deck, the subject can be rewarded or punished in terms of fictive money. Some of the decks are advantageous in the long run, i.e. the 7

reward is greater than the punishment, and some are disadvantageous. Before starting the IGT performance, the subject is instructed to play to increase profit (18). First the IGT was used for patients with impairment in the ventromedial prefrontal cortex (18) but, over time, the task has also been used for research regarding other issues, such as drug abuse and gambling behaviour. In everyday life people have to make complex choices. The IGT test is suggested to capture the complexity of decision-making. Accordingly, if a person wants to do well in the test the subject has to use a mixture of various executive abilities (9). Healthy people playing the IGT test typically pick cards from all decks in the beginning of the performance. After a while a shift from an explorative to a more strategic approach, where the advantageous decks are selected more frequently, is usually seen. A patient with a prefrontal brain damage that performs the IGT selects cards from all decks in the first card selections, and thereafter returns to choosing more bad decks (18). Further conditions associated with poorer IGT performances are alcohol abuse, drug addiction and increased risk behaviour (19). 1.5 Combined oral contraception It was more than 50 years ago that the combined oral contraceptive (COC) for women was released on the market (20). According to information from the Swedish Medical Products Agency in 2005, 75% of the women in Sweden used contraception during their previous sexual intercourse. Further, the most common contraceptive method was hormonal contraception (21). Additional studies suggest that the combined oral contraceptive pill is one of the most frequently used contraceptives among young women in the US and in Western Europe (22) (23). Moreover, many women experience improvements in mood while taking treatment (24), and combined hormonal contraceptives can be used to treat severe premenstrual symptoms, i.e. premenstrual dysphoric disorder (PMDD) (25). According to the results of the randomised controlled trial used for this thesis, mood improvements, in terms of decreased depression, was noted in the premenstrual phase among combined oral contraceptive users (26). Yet, combined oral contraception is also associated with side effects. Mood issues, such as symptoms of depression, mood swings, irritability (24) and the fear of mood effects (21), are common reasons for discontinuing the treatment (24) (21). Also, side effects in terms of experienced weight gain and problems with bleedings are common reasons for discontinuing the contraceptive (21). Furthermore, some women seem to be more sensitive to negative side-effects caused by combined oral contraceptives, and mood side effects are shown to be more prominent in the intermenstrual phase of the treatment cycle (26). 8

There are many studies about how the menstrual cycle is linked to different types of impulsive behaviour among women. In a study by Protopescu and colleagues, neuroimaging findings demonstrate that the activity of the OFC is affected by the menstrual phase (27). This is also supported in another study where it is seen that inhibitory control fluctuates across the menstrual cycle (28). However, few studies have investigated the impact of hormonal contraception on emotional processing and cognitive functions in the brain. In addition, most previous studies have been cross-sectional, opening up for selection bias. A previous, small, placebo-controlled combined oral contraceptive study investigated the brain activity during response inhibition, (20). According to the study (20), oestrogen and progesterone are implied to affect prefrontal dopaminergic function, and therefore cognitive control should also be affected. The Go/No Go tasks, which was used in the study by Gignell and colleagues, measures the ability to avoid learnt behaviour reactions (20). If the inhibitions in these tasks are accurate, networks in the dorsal medial frontal cortex, the lateral frontal cortex and the insula are triggered. However, the study results imply that the women randomised to combined oral contraceptives had a minimal positive impact on brain activity during response inhibition (20). As for the combined oral contraceptive pill Zoely, it contains hormones in terms of 1.5 mg estradiol (E2) and 2.5 mg nomegestrol acetate (NOMAC) (26) (29). Moreover, the NOMAC/E2 contraceptive is proposed to have a very low impact on mood due to the 24/4 regimen (29) (26) and the high content of progesterone-selective progestagen (26). Additionally, women taking NOMAC/E2 are reported to experience an improvement in Negative Affect (loneliness, anxiety, mood swings, crying, irritability, tension, feeling sad or blue and restlessness), Behaviour Change (poor school or work performance, take naps or stay in bed, stay at home, avoid social activities and decreased efficiency), Impaired Concentration (insomnia, forgetfulness, confusion, poor judgment, difficulty concentrating, distractible, minor accidents and poor motor coordination), Pain and Water Retention during premenstrual and menstrual phases (29). Cloninger has described inherited personality traits of an individual through three independent and interrelating elements; novelty seeking, harm avoidance and reward dependence (30). Novelty seeking includes the tendency of looking for new events and a strong sense of thrill to new stimuli. Harm avoidance comprises the drift to react powerful to aversive stimuli and to avoid penalty as well as new events and non-reward. As for reward dependence, it includes intense respond to rewards and the continuance of rewarded behaviour (30) (31). The mixture of high novelty seeking, high reward dependence and low harm avoidance is a dangerous combination, it can result in 9

critical situations and is associated with long-term anxiety (30). Subsequently, increased risk behaviour can be potentially dangerous and is important to identify. As Zoely is suggested to have a minimal impact on mood compared to other combined hormonal contraceptives (26) (29), a demonstrated change in risk behaviour with Zoely might also be applicable among other combined oral contraceptives. Depending on the results, this sub-study could either help to disprove earlier research about a slight difference in impulsivity between combined oral contraceptive users and non-users (20), or it could be helpful in identifying women that can experience behavioural improvements by taking combined oral contraceptives. If the latter, it might affect future indications for prescribing combined oral contraceptives. 1.6 Research Question There are several studies about impulsive behaviour among women across the menstrual cycle and variations in inhibitory control have been demonstrated (27) (28) (20). However, when searching in PubMed for RCT studies about how combined oral contraception affects risk behaviour, no studies were found. According to previous findings, the oral contraceptive used for this study has been associated with enhanced judgement and concentration (29). In this RCT study we aim to investigate our hypothesis that women on combined oral contraceptive display decreased general risk behaviour in comparison to women without hormonal contraception. To investigate whether risk behaviour is affected, statistical analyses of Iowa Gambling Task performances are conducted. 10

2 Material and method 2.1 Participants The original study, upon which this thesis rests (26), started as an answer to the absence of RCT studies regarding combined oral contraceptive use and mood issues. The study aimed to investigate unfavourable mood effects in combined oral contraceptive users in a study population that was as unselected as possible. The project ranged between 7 September, 2013 and 29 September, 2015 and took place at the Departments of Obstetrics and Gynecology at Uppsala University Hospital, Södersjukhuset, Karolinska University Hospital, Umeå University Hospital, Linköping University Hospital, Örebro University Hospital and Närhälsan Maternity Health Care Center in Frölunda, Gothenburg (26). In total, 202 healthy women aged 18-35 years were included in the study (26). However, the psychological tasks which will be investigated in this thesis, were only carried out in Uppsala. Notices in local papers, websites for students and local boards helped recruit the participants. Inclusion criteria, besides age, were body mass index below 30 kg/m 2 and the will to use backup contraception throughout the entire study period. The only exclusion criteria in the study were the usual exclusion criteria for combined oral contraceptive prescription exerted in clinical practice; a history of venous thromboembolism in the family, two or more risk factors for venous thromboembolism, identified thrombophilia, identified dyslipidemia, systolic blood pressure > 140 mmhg or diastolic blood pressure > 90 mmhg, first degree relatives with cardiovascular disorder at early age, migraine with focal symptoms, inflammatory disorders, former cancer, liver disorders, former pancreatitis and the use of treatments which can compromise the uptake or metabolism of the combined oral contraceptive (26). To gain further knowledge about the participants, a screening visit was conducted. In addition to age and BMI, information about any children, previous abortions and years in school was obtained. Furthermore, standardized interviews were performed to gain information about earlier combined oral contraceptive use among the participants. Women who had never taken any hormonal contraceptive prior to the study were referred to as never users, while women who had started taking a contraception but then, for various reasons, stopped were called starters. Switchers were the women who had another hormonal contraception prior to the study. During the screening visit, trait impulsivity was estimated with the Swedish universities Scale of Personalities (SSP) (32), smoking and snuff habits were asked for and risk use of alcohol was obtained with Alcohol Use Disorders Identification Test (AUDIT) (33). Additionally, all women filled out the self-rated 11

version of the Montgomery-Åsberg Depression Rating Scale (MADRS-S) during the screening visit (26). The MADRS-S is mainly a tool to notice change in depressive symptoms (34), however, in Swedish primary care, a score of 15 or above indicates current depression (26). Moreover, it was important to identify signals of premenstrual syndrome (PMS) and PMDD. Hence, the Daily Record of Severity of Problems (DRSP) was administered and filled-out during the baseline cycle, i.e. one menstrual cycle before treatment where the women were not allowed to take any hormonal contraception (26). The DRSP is a well-known tool to diagnose and follow up women with PMDD and measure the severity of symptoms during different periods of the menstrual cycle (35). All in all, the participants were well informed about the procedure as well as the aim of the study and signed a printed informed agreement (26). 12

Figure 1 exhibits a flow chart of the participants. One hundred out of 202 women were included in Stockholm, Gothenburg, Örebro, Linköping and Umeå, whereas 102 participants were included in Uppsala. As regards the 102 women in Uppsala, sixteen dropped out, four did not want to attend and 42 participants were included in other psychological studies prior to this sub-study and did not want to attend in a further study. In total, forty women were included in this sub-study, aiming to investigate risk behaviour among women, using combined oral contraceptive, through statistical analyses of the IGT (Figure 1). The study was acknowledged by the Independent Ethical Review Board at Uppsala University, Sweden, and implemented at the Department of Women s and Children s Health at Uppsala University Hospital in Sweden. The study identifier is Dnr 2013/161. 202 Total number of participants 100 Number of participants in Stockholm, Gothenburg, Örebro, Linköping and Umeå 102 Number of participants in Uppsala 16 Dropouts 4 Participants not willing to attend 42 Participants included prior to substudy 40 Participants included in sub-study Figure 1. Flow chart of participants. 13

2.2 Study design The study is designed as an investigator-initiated, double-blinded, placebo-controlled randomised trial. After the baseline cycle, all participants were randomised to be treated with either combined oral contraceptive in terms of Zoely (1.5 mg estradiol and 2.5 mg nomegestrolacetate) or placebo throughout three 24/4 treatment cycles. Consequently, every treatment cycle started with 24 days of taking one daily pill and ended with four days without tablets. The randomisation, including the packing, was conducted by Apoteksbolaget Production and Laboratories (APL, the National Corporation of Swedish Pharmacies). Neither the patient nor the study employees knew about which pill the participant received and the randomisation codes were stored at the Uppsala University Hospital Pharmacy. When all three treatment cycles were completed, the participants returned with their DRSP ratings and the pill packages. Moreover, a study nurse counted any remaining pill (26). The participants in the study filled out the DRSP during the baseline cycle and in the last treatment cycle. Additionally, all women filled out the self-rated version of the Montgomery-Åsberg Depression Rating Scale (MADRS-S) during both the screening visit and the last treatment visit (26). Finally, during the last treatment visit in Uppsala, the women were asked to do a psychological task, one of which was the IGT. The women that chose to do the IGT was included in this study (Figure 1). 2.3 Iowa Gambling Task The women who were included in the sub-study performed the IGT (Figure 2) at the last treatment visit and thus after the three treatment cycles. The women were placed in front of a computer screen, facing four decks of cards (A, B, C and D), all appearing identical in size and feature. Further, the decks were displayed in randomised order on the screen. The participants turned the cards from any deck, one at a time, and were allowed to change deck whenever they wanted to and as many times as they desired (18). All in all, each participant made 100 picks during the IGT. Before starting the test, all participants received a loan of fictive money, 2,000 i-dollars ($ i ), on the computer screen to start with. They also obtained the same written instructions to play to increase their profit. The instructions in Swedish were the following; In this test, you will repeatedly be able to pick a card from any of the four decks. You choose deck by clicking on the button beneath the deck with your mouse. For 14

every picked card, you can be rewarded in i-dollars, but you can also lose some. You will get 100 picks in total to maximize your profit. The present earned amount of money and the performed number of picks will be visible on the screen after every card selection. The trick is that some of the decks will be more advantageous than others. Try to select cards from the most advantageous decks. You will start with 2000 $ i. When choosing decks C and D, the subjects gained a small amount of money, 50 $ i, if they were rewarded. However, if they lost money the penalty was also small. The punishment was only 50 $ i in deck C and 250 $ i, but less frequent, in deck D. Overall, the subjects gained more money than they lost if they picked only these decks. Decks C and D were therefore profitable in the long run and are thus called advantageous decks. When selecting decks A and B, the reward was greater, 100 $ i, compared to deck C and D. The penalty, on the other hand, was also higher in these highpaying decks. If the participants lost money, the punishment ranged between 150 to 350 $ i when choosing deck A. In deck B, penalty was not as frequent as in deck A, but when it occurred it could be as high as 1,250 $ i. The punishment exceeded the reward when the participants selected these disadvantageous decks A and B (18). Disadvantageous Advantageous Gain: 100 $ i A B C D Gain: 100 $ i Gain: 50$ i Gain: 50 $ i Loss: 150 350 $ i Loss: 1,250 $ i Loss: 50 $ i Loss: 250 $ i Figure 2. Iowa Gambling Task, a schematic view. 2.4 Statistical analyses As for the randomisation procedure, the women were randomised to either Zoely or placebo in the original study by Lundin and colleagues (26). Thereafter, some of the women in Uppsala agreed to do a psychological task. No formal power analyses were conducted for this sub-study as it was no 15

primary aim for the trial. Regarding the statistical analyses, the calculations were performed using the program R. Starting with the participant characteristics, the variables presented aimed to reflect basic information about the two groups, as well as it intended to reflect issues that may have an impact on risk behaviour. Differences between treatment groups were compared by Mann-Whitney U tests, Chi-square tests, Fisher s Exact tests and Independent Samples T-tests. As for the IGT analyses, the first statistics were administered to obtain knowledge about how all participants behaved at a group level when performing the IGT. Primary, the distribution of chosen decks amongst all participants were analysed to gain an overview of which decks the 40 women preferred. The calculations were performed in median and interquartile range (IQR) and the P- values were analysed with Wilcoxon Signed Rank tests. Thereafter the overall performances throughout the IGT were analysed to investigate the distribution between selected advantageous and disadvantageous decks over time. The hundred picks were divided into five quintiles, each containing twenty consecutive selections. Accordingly, this partition made it possible to follow the results over time during the test. Median and IQR as well as Wilcoxon Signed Rank tests were administered. As for the IGT statistics regarding results depending on the treatment (i.e. combined oral contraceptive or placebo), the distribution of chosen decks in the two groups was first analysed, by median, IQR and Mann-Whitney U tests. Thereafter, the proportions of chosen advantageous as well as disadvantageous decks throughout the IGT were analysed. As the selections over time were of interest, division into quintiles was conducted. Further, median, IQR and Mann-Whitney U tests were applied. Finally, the overall performances over time were analysed separately in both groups in order to discover if any shift in preference for advantageous decks could be distinguished. Median and mean were applied in these later analyses. Moreover, an omnibus three-way ANOVA was made to discover differences between the groups, adjusting for the variable years in school. Additionally, the eventual presence of linear interactions in the respective group was investigated through repeated measures ANOVA. All the statistics using ANOVA was conducted in SPSS. Regarding the total financial outcome, mean and standard deviation (SD) as well as independent samples T-tests were exerted. As for all the statistics, the Shapiro-Wilk normality test, as well as observations of histograms, were performed to investigate whether the data were normally distributed or not. Additionally, only statistical significant P-values (p < 0.05) were displayed in the figures. 16

3 Results 3.1 Subjects All in all, 40 of the women included in the original RCT study by Lundin and colleagues participated in the sub-study (26). Twenty-two of these women who agreed to attend were randomised to the combined oral contraceptive and 18 women to placebo. Participant characteristics are displayed in table 1. The two groups did not differ concerning age, weight, women with children, previous induced abortion and hormonal contraceptive use prior to inclusion (starter, switcher and never-user). Furthermore, no differences between the groups were noted regarding trait impulsivity, smoking, snuff use, alcohol risk use, MADRS-S before and after treatment and DRSP changes from the baseline cycle to the last treatment cycle. As regards education, women in the combined oral contraceptive group had, in average, completed 14.5 years in school whereas women in the placebo group had completed 16.1 years (p = 0.022) (Table 1). 17

Table 1. Participant characteristics in the combined oral contraceptive and the placebo group. Variable Combined oral contraceptive (n=22) mean (SD), median (IQR) or n(%) Placebo (n=18) mean (SD), median (IQR) or n(%) p-value Age, years, median (IQR) 22.5 (4.5) 25.5 (4.0) 0.115 Age <25 years, n (%) 16 (72.7) 7 (38.9) 0.067 Normal weight, BMI (kg/m 2 ) 18.5-24.9, (%) 19 (86.4) 15 (83.3) 1.000 Overweight, BMI (kg/m 2 ) 25.0-29.9 (%) 3 (13.6) 3 (16.7) Women with children (%) 1 (4.5) 1 (5.6) 1.000 Previous induced abortion (%) 3 (13.6) 0 (0.0) 0.239 Years in school, mean (SD) 14.5 (1.3) 16.1 (2.3) 0.022 Hormonal contraceptive use prior to inclusion: Starter (%) 14 (63.6) 12 (66.7) 0.809 Switcher (%) 4 (18.2) 2 (11.1) Never-user (%) 4 (18.2) 4 (22.2) Factors associated with increased risk behaviour: Trait impulsivity, mean (SD) 48.3 (9.4) 48.1 (9.3) 0.960 Smoking (%) 3 (13.6) 1 (5.6) 0.613 Snuff user (%) 1 (4.5) 1 (5.6) 1.000 AUDIT score, risk use (%) 7 (31.8) 3 (16.7) 0.465 Before treatment, MADRS-S, median (IQR) 5.5 (2.3) 3.5 (5.5) 0.411 After treatment, MADRS-S, mean (SD) 7.4 (4.1) 8.2 (5.5) 0.631 DRSP, 20-40% improvement (%) 1 (4.5) 0 (0.0) 1.000 DRSP, 20-40% deterioration (%) 4 (18.2) 3 (16.7) DRSP, 0-20% improvement (%) 9 (40.9) 8 (44.4) DRSP, 0-20% deterioration (%) 8 (36.4) 7 (38.9) BMI = Body Mass Index, MADRS-S = The self-rated version of Montgomery-Åsberg Depression Rating Scale, DRSP = Daily Record of Severity of Problems, IQR = Interquartile Range, SD = Standard Deviation 18

3.2 Iowa Gambling Task Altogether, the most popular decks amongst the 40 participants were B and D (Figure 3). As regards advantageous and disadvantageous decks, the participants chose deck A and deck B to the same extent as C and D (Figure 3). Figure 3. Distribution of chosen decks amongst all participants. Advantageous decks C and D, disadvantageous decks A and B. Performances throughout the IGT demonstrated that the participants selected the disadvantageous decks to a greater extent than the advantageous decks at the beginning of the task, i.e. in block 1 and 2. In the last three blocks, deck A and B were chosen as often as deck C and D (Figure 4) Figure 4. Overall performances throughout the IGT. 19

3.3 Outcome depending on treatment Further to the IGT results depending on treatment. As demonstrated in figure 3, decks B and D were the most popular. In figure 5 it is notable that this finding is applicable in both the combined oral contraceptive and the placebo group. As regards the distribution of chosen decks, no differences could be seen between the groups (Figure 5). Figure 5. Distribution of chosen decks in the combined oral contraceptive and the placebo group. Advantageous decks C and D, disadvantageous decks A and B. Figure 6 exhibits to what extent the advantageous decks, C and D, were chosen in the combined oral contraceptive and the placebo group throughout the IGT. During block 2, the participants randomised to combined oral contraceptive selected more advantageous decks compared to the women on placebo. As for the four remaining blocks, no statistical significant difference was noted (Figure 6). Additionally, figure 7 displays the frequency of chosen disadvantageous decks, A and B, in the two groups during the IGT. A statistical significant difference is notable during block 2, where the women on combined oral contraceptive picked less disadvantageous decks. However, as in figure 6, no statistical significant difference could be distinguished during the other four blocks (Figure 7). 20

Figure 6. The proportion of advantageous decks chosen throughout the IGT in the combined oral contraceptive and the placebo group. Figure 7. The proportion of disadvantageous decks chosen throughout the IGT in the combined oral contraceptive and the placebo group. Overall performance in relation to randomised treatment was evaluated in an omnibus three-way ANOVA (deck x time-point x treatment), adjusting for years of education. No difference between treatments was noted as evidenced by a non-significant three-way interaction F (1, 37) = 1.44; p = 0.238. However, because of prior reports of improved judgement and concentration in combined oral contraceptive users, we pursued the ANOVA analyses in each separate group. As for the overall performances in the combined oral contraceptive group (Figure 8 and 9), the 22 women included in this group picked the disadvantageous decks more frequently in the beginning of the task, i.e. in block 1 and 2 (Figure 8 and 9). Yet, as the IGT continued, a shift to a more strategical approach, 21

where a majority of advantageous decks were selected, was seen according to the parametric statistics (Figure 9). This shift occurred just prior to block 3 and lasted during the three last blocks. Repeated measures ANOVA revealed a linear interaction F (1, 21) = 5.28; p = 0.032. Figure 8 (the diagram to the left) and 9 (the diagram to the right). Overall performances in the combined oral contraceptive group according to non-parametric statistics in terms of median, figure 8, and parametric statistics in terms of mean, figure 9. The overall performances in the placebo group are displayed in figure 10 and 11. The 18 women randomised to placebo selected the disadvantageous decks more often in the beginning of the IGT, and continued to do so throughout the test (Figure 10 and 11). Thus, no shift in preference for advantageous decks was noted, as evidenced by a non-significant linear interaction in the ANOVA, F (1, 17) = 2.32; p = 0.146. Figure 10 (the diagram to the left) and 11 (the diagram to the right). Overall performances in the placebo group according to non-parametric statistics in terms of median, figure 10, and parametric statistics in terms of mean, figure 11. 22

Finally, table 2 reveals the total financial outcome in the oral contraceptive and the placebo group. On the whole, the women randomised to either of the groups gained less than they lost, resulting in overall negative end results. Moreover, there were no dissimilarities between the groups concerning total gain, total loss and summation (Table 2). Table 2. Total financial outcome. Variable Contraceptive mean (SD) Placebo mean (SD) P-value Total gain 7,530 (893) 7,839 (673) 0.220 Total loss 7,827 (1,792) 8,189 (1,359) 0.473 Summation -298 (1,021) -350 (782) 0.856 23

4 Discussion 4.1 Introduction, method and results As earlier mentioned, the IGT is a popular tool to measure risk behaviour (16). In this sub-study, we aimed to investigate whether women on combined oral contraceptive display decreased risk behaviour, compared to women on placebo, through analyses of IGT results. The main findings in this report were that all participants showed a preference for decks B and D and an overall even distribution between picked advantageous and disadvantageous decks was noted. As regards results over time, the participants preferred disadvantageous decks during the first two blocks, thereafter the disadvantageous decks were picked as often as the advantageous. Concerning results depending on treatment, the participants chose the disadvantageous and the advantageous decks to the same extent during all blocks except for block 2, where the women randomised to combined oral contraception picked more advantageous decks. Additionally, unlike the placebo group, a linear interaction was seen in the combined oral contraceptive group prior to the third block, verifying a shift in preference for advantageous decks over time. As for the total financial outcome, it was likewise negative in both groups. Accordingly, the participants in the study favoured decks B and D, where the penalty was not as numerous (18). A preference for these non-frequently punishing decks has also been notable among both healthy decision makers and patient groups in other studies, where the fondness for deck B, regardless of negative long-term outcomes, is referred to as prominent deck B (36). This suggests that the participants preferred frequent gains, rather than profitable long-term outcomes (37) (38). Further, this idea is supported by overall negative end results in both groups demonstrated in table 2. Moreover, it was also seen that the women picked more disadvantageous decks in the first two blocks, but then started to pick the advantageous decks to the same extent as the disadvantageous decks during the remaining of the IGT performance. This indicates that the participants understood the task, at least to some degree, and avoided disadvantageous decks more, but not entirely, during the last three remaining blocks. However, the results discussed refer to both women randomised to combined oral contraceptive and to placebo. When observing the groups separately, a shift in preference for advantageous decks is only seen in the combined oral contraceptive group. Consequently, according to the parametric statistics, the women on combined oral contraceptive made a shift from an explorative approach to a strategical approach to maximize long-term profits, as usually seen among healthy subjects when performing the IGT (18). The shift in preference occurred just prior to the third block. Regarding the difference in picked advantageous 24

and disadvantageous decks between the groups, it was found in the second block. Accordingly, the women on placebo continued with an explorative approach, while the women on combined oral contraception started to choose long-time profitable decks. Further, the shift absence among the women on placebo implies that these participants continued to operate a more explorative approach during the entire IGT. The failure to avoid disadvantageous decks is suggested to be explained by an insensitivity to forthcoming consequences (18) and, as earlier mentioned, the preference for frequent gains instead of profitable outcome over time is observed in several studies (37) (38). Still, the participants in the placebo group were healthy young women and a shift in preference for advantageous decks was anticipated (18). As for the women randomised to combined oral contraception, in order to make the preference change, the participants resisted temporary high reward in the disadvantageous decks, thus requiring behaviour control (9). Moving to the study population, the number of participants in the study was small, with only 40 participants in total. Therefore, the risk for random outcomes is evident and it is difficult to discover differences between the groups that might be true, in statistics referred to as a type II error (39). As for the Iowa Gambling Task, the IGT is a suitable and popular, but not a perfect, task and long-run profits is not the only factor that influence the participants choices (38). The IGT measures risk behaviour, which is a complex construct, and there is an increased risk for random outcomes. As for all neuropsychological instruments, the IGT results must be understood in combination with other aspects. For example, present mood, as well as the working memory and the personality of the performer, also influences the results (40). To conclude, personality characteristics and state mode must be evaluated continuously when interpreting the IGT results. In the sub-study, personality in terms of trait impulsivity and other personal characteristics were estimated prior to the study. Regarding current mood, DRSP and MADRS-S were assessed in the baseline cycle and in the last treatment cycle. Regardless of the time point, neither trait impulsivity, DRSP ratings nor MADRS-S scores differed between the groups. Continuing with the lack of strategy shift among the women on placebo. There might be several explanations to the shift absence, one of which is random diversities between the groups. In table 1, the participant characteristics are displayed, and only the variable years in school differs. The women randomised to placebo had completed 16.1 years in school, whereas the women randomised to combined oral contraceptive had finish 14.5 years in mean (p= 0.022). As higher education is associated with higher IQ (14), the women in the placebo group might have been smarter than the women in the combined oral contraceptive group, thus smoothing out eventual differences among 25

the results depending on treatment. However, no IQ tests were performed and the shift in the combined oral contraceptive group, and the absence of strategy change in the placebo group, rather supports that the women in the combined oral contraceptive group had the intellectual benefit. However, an adjustment for years of education was made with an omnibus three-way ANOVA to exclude eventual deceptive results, and there were no statistical significant dissimilarities between the groups. In addition to the described variance in participant characteristics, the two groups were equivalent, including factors associated with increased risk behaviour and trait impulsivity (Table 1). As for the age variable in table 1, the frontal lobes are highly involved in decision making (8), and a fully developed brain should be favourable when performing the IGT. In a previous study, the performance of the IGT was evaluated in healthy girls and boys aged 9-17 years (19). It was found that the participants within the age range of 14 to 17 years picked more advantageous decks and also made a shift in preference earlier during the performance, compared to the youngest participants. Yet, the adolescents aged 14-17 years did not reach the same amount of positive scores from the advantageous decks as seen among healthy adults performing the IGT in other studies (19). This finding supports that young age is a disadvantage when performing the IGT. The median age in the placebo group was 25.5 years, whereas it was 22.5 years in the combined oral contraceptive group. Regardless of a non-statistical significance between the groups in median age (p = 0.115) and age category (p =0.067), many cognitive changes occur during the adolescence and there might be cognitive differences only in a few years, thus in favour for the older participants in the placebo group. As for the frontal lobes, the brain is not fully developed until around the age of 25 (15) and it is said that the frontal lobes mature first during late adolescence (13). When comparing two groups, many aspects need to be considered to make a proper comparison, and there is always a risk that important variables remain unexplored. For example, it is explained that lower leptin, rather than lower body fat and BMI, which was used in this study, is associated with better IGT performances (41). Additionally, stress also interferes with the achievements in the IGT and high cortisol levels are reported to be linked to worse performances (42). In the sub-study, no cortisol levels nor leptin levels were measured. Additional factors not taken into consideration in this sub-study were further addictive behaviours besides smoking habits, snuff use and alcohol habits. Poor impulsive performances in the IGT have been noted among people with gambling disorder, polysubstance dependent individuals and stimulant abusers (43). Furthermore, the adolescence is a period in life when people are more risky-taking (13) (15), impulsive and novelty 26

seeking (13). Moreover, combined oral contraceptives contains hormones and no quantifying of sex hormones was executed. According to a study by Gingnell and colleagues, (20), oestrogen and progesterone are proposed to affect cognitive control by the influence of prefrontal dopaminergic function. However, earlier studies demonstrate diverse results and a negative linkage between estradiol levels and sensation seeking is also described in some research (44). As regards the participant characteristics in the study, even if not statistically significant, the women randomised to combined oral contraception were more often smokers and risk users of alcohol, as well as they were a few years younger. Considering these non-significant possibly disadvantageous units, they rather imply that the women randomised to combined oral contraceptive were slightly riskier taking. Still, these women performed the IGT in a more advantageous way. Even if it would have been suitable to take additional addictive behaviours into consideration and the measuring of sex hormones would have added interesting information, these actions would probably not be essential in explaining our results in this study. As for other research, improvements in Negative Affect (29), Impaired Concentration (29) and overall quality of life (25), has been noticed among combined oral contraceptive users (45). Regarding hormonal contraception, it has been shown that oestrogen modulates 5-HT and norepinephrine (NE), both important in cognitive adaption and mood (46). As earlier mentioned, dopaminergic function, and therefore cognitive control, is also suggested to be affected by oestrogen (20). In the sub-study, the women randomised to combined oral contraception obtained Zoley. The combined hormonal contraception contains E2 and the supply of E2 contributes to an increased level of 5-HT, as well as it increases the amount of 5-HT receptors in important limbic brain regions (46). Even if it is probable that decision-making is affected by better concentration and improved judgement, the studies are not designed to evaluate risk behaviour. When searching in PubMed for randomised controlled trials (RCT), treating the role of combined oral contraception on risk behaviour, no articles about the topic is found. According to the study by Gignell and colleagues, most studies regarding combined oral contraceptives and their influence on the brain contain a cross-sectional design and combined oral contraceptive users are compared to non-users (20). Obviously, there is a risk that women that choose to take combined oral contraception in the first place are less risky-taking to begin with. As for this sub-study, it was an investigator-initiated, double-blinded, placebo-controlled randomised trial. Subsequently, even if there is a risk that the groups were not comparable and that true differences might have passed unnoticed because of the small number of participants, the groups were as equivalent as possible and the treatment was based on randomisation. 27

Overall, the results among the women in the respective groups were similar. Yet, the women randomised to placebo never made a shift in preference for positive long-run outcome, and the most likely explanation is that they did not understand the test as fast as the women randomised to combined oral contraceptive. It is expected that, eventually, a shift would have appeared even in the placebo group. To summarize, the presence of the positive preference shift in the combined oral contraceptive group, and the absence of it in the placebo group, may imply that the women in the combined oral contraceptive group were more intelligent. However, the results are ambiguous and this prevents us from drawing well-founded conclusions. Accordingly, further studies are needed to investigate whether combined oral contraceptives affects risk behaviour. 4.2 Benefits and limitations By comparison, most studies on cognitive tests in hormonal contraceptive users have been crosssectional, i.e. investigating users with non-users (20). Needless to say, such design opens up for selection bias, and are difficult to interpret. Starting with the benefits of the sub-study, as demonstrated in the main study (26), it is an investigator-initiated, double-blinded, placebocontrolled randomised trial. Consequently, neither the personnel in the study nor the participants knew what treatment each patient was given, which increases the reliability of the results. Additionally, the study population was selected as little as possible in order to make a truthful representation of young women, and a new finding in the report would therefore be more likely to be applicable among other women outside of the study. Furthermore, when observing participant characteristics, no major differences were seen among the groups, which indicates that the participants were as comparable as possible in the sub-study. Additionally, state modes in terms of MADRS and DRSP were evaluated two times during the study. As regards the actual performance of the IGT, all participants received the same written instructions on the computer screen prior to the start of the performance and the task ended after exactly 100 completed deck selections. Also, it is observed that participants improve their results when performing the IGT several times, and the results of repeated performances are not sure to be reliable (40). In the sub-study, the IGT was conducted only once. However, the sub-study also has its limits. To begin with, the study population was small, with only 40 women in Uppsala included. The small number of participants makes it difficult to see trends and reach statistical significance, and the risk for a statistical type II error is evident. Also, the IGT 28