A prospective study of learning, memory, and executive function in new MDMA users

Abstract

Aims

It is still unclear if cognitive abnormalities in human 3,4-methylenedioxymeth-amphetamine (MDMA) users existed before the beginning of use or if other confounders could explain the deficits. The present study was conducted in order to assess the relationship between beginning MDMA use and subsequent cognitive performance and to overcome previous methodological shortcomings.

Design

A prospective cohort study in new MDMA users between 2006 and 2009 with a follow-up duration of 12 months.

Setting and Participants

Of the 149 almost MDMA-naive subjects examined at the initial assessment, 109 subjects participated again after 1 year. During this period, 43 subjects did not use any other illicit substance apart from cannabis; 23 subjects used more than 10 pills MDMA (mean = 33.6). These groups then were compared by means of multivariate analyses of variance.

Measurements

Change scores between the initial examination and follow-up on a neuropsychological test battery including measures of learning, memory, and frontal executive functions [Auditiv-Verbaler Lerntest (AVLT), Lern- und Gedächtnistest (LGT) 3, digit span test, digit symbol test, Stroop task, Trail-making test]. In addition, a comprehensive number of possibly relevant confounders including age, general intelligence, cannabis use, alcohol use, cigarette use, medical treatment, participation in sports, nutrition, sleep patterns and subjective wellbeing was assessed.

Findings

Groups did not differ in any of the potential confounders. However, significant effects of immediate and delayed recall of a visual paired associates learning task between MDMA users and controls were found (respectively, F _(1,64) = 11.43, P = 0.001, η² = 0.136 and F _(1,64) = 11.08, P = 0.002, η² = 0.144). No significant differences on the other neuropsychological tests were found.

Conclusions

MDMA appears to impair visual paired associates learning in new users, suggesting serotonergic dysfunction in hippocampal regions as a consequence of MDMA use.

Introduction

3,4-Methylenedioxymethamphetamine (MDMA), commonly referred to as ‘ecstasy’, is a psychostimulant drug which is popular among young adults. In Europe, the life-time prevalence of MDMA use is 5.6% in the 15–34-year-old population, with highest prevalence estimates in the United Kingdom (12.7%) [1]. For North America the life-time prevalence rates are generally estimated higher than in Europe [2]. MDMA is used illegally as a recreational drug, in particular by visitors of techno clubs and rave parties [3].

MDMA and related substances act primarily upon indirect serotonergic and dopaminergic mechanisms in the central nervous systems (CNS). Since the mid-1980s, repeated administrations of MDMA to experimental animals have suggested neurotoxic degeneration of serotonergic axon terminals followed by a decrease of 5-HT (5-hydroxytryptamine; serotonin) concentration in brain tissue [3]. This effect persists even after several years [4, 5]. Studies investigating different markers of 5-HT [main metabolite in cerebral spinal fluid (5-hydroxyindoleacetic acid), 5-HT transporter density, postsynaptic 5-HT receptors] in human users, support the hypothesis that the suggested neurotoxic potential of MDMA which was reported in laboratory animal research could be relevant to humans [5-10]. With this it is to be considered that, in contrast to the neurodegeneration hypotheses, neuroregulatory mechanisms are suggested to underlie MDMA-induced serotonergic dysfunction [11]. 5-HT is involved in many functional systems, including psychopathology, neuroendocrine and sleep regulation as well as regulation of vegetative functions. Additionally, a number of impairments in cognition and stimulus processing are conceivable as a consequence of MDMA-induced serotonergic dysfunction.

With regard to cognitive performance, a large number of cross-sectional studies and a handful of longitudinal investigations indicate that the most consistent findings have been decrements in memory and learning performance [12]. However, deficits in working memory, planning ability and central executive control, as well as high cognitive impulsivity, have also been reported [3, 13, 14].

In general, most studies on the cognitive effects of MDMA suffer from several methodological problems (e.g. pre-existing differences, polydrug use, differences in life-style). Therefore, a prospective methodological approach was applied in order to assess cognitive performance. In the present study, a comprehensive neuropsychological test battery was applied to assess cognitive performance over the course of 1 year in new MDMA users. In addition, a comprehensive number of possibly confounding variables including age, general intelligence, cannabis use, alcohol use, cigarette use, medical treatment, participation in sports, nutrition, sleep patterns and subjective wellbeing were explored and controlled for with regard to the statistical analyses. The goal was to address the following questions.

• Does the use of MDMA over a period of 1 year lead to a cognitive performance decrease?
• Which cognitive domains are most vulnerable to be affected by initial/incipient use of MDMA over a period of 1 year?
• Are the potential effects of initial/incipient MDMA use on cognitive impairments confounded by age, general intelligence, cannabis use, alcohol use, cigarette use, medical treatment, participation in sports, nutrition, sleep patterns and/or subjective wellbeing?

Method

Participants

One hundred and forty-nine new MDMA users with no current physical disorder and no current or previous history of neurological or psychiatric disorder (Axes I and II according to DSM-IV criteria; APA, 1994) were included in the study. Further exclusion criteria were the following: ingestion of any other illicit psychotropic substances besides cannabis on more than five occasions before the day of the first examination; a history of alcohol misuse (according to DSM-IV criteria, APA 1994); and regular medication (except for contraceptives). Main inclusion criterion at baseline was a high probability of future ecstasy use, operationalized as having first but very limited experience with MDMA (maximum five pills). After 12 months participants were invited back. Of the initial 149 subjects, 109 subjects [72 males, 37 females; age range at baseline: 18–35 years, mean: 23.42 years, standard deviation (SD) 4.76] participated in the second assessment. Cognitive assessment was carried out when participants were abstinent from cannabis on both study days in order to rule out acute intoxication effects. Given the fact that most MDMA users also use cannabis it would have been implausible to recruit MDMA users with a longer period of abstinence [15]. Furthermore, participants had to be abstinent from any other illicit substance for at least 7 days in order to rule out acute intoxication effects. Subjects were recruited via advertisements in magazines and newspapers and via notifications posted on campus. The study was part of a larger investigation including psychopathological and neuroimaging measures that will be submitted elsewhere. The study was approved by the Ethics Committee of the Medical Faculty of the University of Cologne.

Procedure

Preceding the cognitive assessment, written informed consent was given by all participants. This was followed by a structured interview developed in order to assess the use of illicit psychotropic substances. For all prevalent psychotropic substances, the interview included questions concerning the age of first use, the number of days since the last use, the average and maximal frequency of use measured in days per month, the estimated cumulative life-time dose, the average daily dose, the highest daily dose ever used and the duration of regular use measured in months [for the second assessment, the interview included questions concerning the following criteria: the age of first use (only assessed if the relevant substance had not been used before), the number of days since the last use, the average and maximal frequency of use measured in days per month last year, the estimated cumulative dose last year, the average daily dose last year, the highest daily dose last year and the duration of regular use measured in months last year]. Qualitative drug screens were performed on the day of the examination by means of urine samples for amphetamines, benzodiazepines, cocaine, methadone, MDMA and cannabis (enzyme-multiplied immunoassay; von Minden GmbH Regensburg, Germany). Furthermore, hair samples were taken randomly in one-third (for economic reasons) of the participants and analysed for the substances MDMA, 3,4-methylenedioxyamphetamine (MDA), 3,4-methylenedioxy-N-ethylamphetamine (MDEA), amphetamine, methamphetamine and cannabinoids by the Institute of Legal Medicine of the University of Cologne in order to verify the self-reported substance use. In addition, a questionnaire regarding health behaviour was used in order to control for confounding variables such as alcohol and cigarette use, sleep patterns, nutrition, participation in sports and subjective wellbeing (Fragebogen zur Erfassung des Gesundheitsverhaltens; FEG) [16].

Neuropsychological test battery

The selection of tests in the present study is based on the results of previous cross-sectional and longitudinal studies of MDMA users, which identified alterations in the areas of working memory, learning, memory and frontal executive functions.

Auditiv-Verbaler Lerntest AVLT

Verbal declarative memory performance was examined by the Auditiv-Verbaler Lerntest (AVLT) [17], which is a German version of the Rey Auditory Verbal Learning Test (RAVLT) [18]. This instrument assesses verbal declarative memory performance by means of immediate recall, total acquisition performance across five trials, recall after interference, loss after interference and recognition after 30 minutes.

Lern- und Gedächtnistest LGT 3

Figural visual recognition was assessed by a subtest of the Lern- und Gedächtnistest (LGT) [19], which is a classical paired associates learning task. The test contains 20 logos, each composed of a central figure and a surrounding frame. These logos were presented to the subjects on a sheet of paper for 60 seconds. Immediately thereafter (immediate recall) and after a delay of 1 hour (without a further learning trial—delayed recall), the central figures of the logos were presented combined with four options of frames. Subjects had to find the correct frame. This test has proved to be sensitive in identifying performance deficits in ecstasy users [9].

Digit-Span-Test

This classical working memory task is part of the Hamburg-Wechsler-Intelligenztest für Erwachsene (HAWIE-R) [20], a German version of the Wechsler Intelligence Test (WAIS) [21]. The experimenter reads out a sequence of digits to the participant, who has to recall the words immediately in reverse order.

Digit symbol test

This speed of information processing test consists of nine digit-symbol pairs (e.g. 1/-,2/┴ … 7/Λ,8/X,9/=) followed by a list of 93 digits. Under each digit the subject is instructed to write down the corresponding symbol as quickly as possible. The number of correct symbols within 90 seconds is measured. The test is also part of the HAWIE-R [20], a German version of the WAIS [21].

Stroop task

In order to measure cognitive interference/inhibition, we used a German paper-and-pencil version of the classical Stroop task (Farbe-Wort-Interferenztest) [22, 23]. This speed performance test consists of three different kinds of stimuli. In the first run, participants read out names of colours appearing in black ink. In the second run, they are instructed to name the given colour of rectangles. In the third run, names of colours appear in a different ink to the colour named, and the task is to name the colour of the ink. For each run participants are instructed to perform the task as quickly and accurately as possible. The experimenter records the times as well as the corrected and uncorrected errors.

Trail-making test

This test measures mental flexibility [24], and consists of two parts. In part A the participant is instructed to connect 25 circles on a sheet of paper using a ballpoint pen. These circles are numbered from 1 to 25, and participants are required to connect the circles in the correct order as quickly as possible. Part B presents both numbers (1–13) and letters (A–L), and requires the participant to connect numbers and letters alternately in the correct order (i.e. 1-A-2-B-3-C). For both parts the response time is recorded.

Raven Standard Progressive Matrices

In order to control for non-verbal general intelligence, the Raven Standard Progressive Matrices [25] were applied. This instrument comprises a series of diagrams or designs, each with a part missing. Subjects were instructed to select the correct part to complete the designs from a number of options printed beneath.

Statistical analyses

Two groups of subjects were defined: those who did not use any other illicit substance apart from cannabis over the course of the 1-year period (non-users) and those who used at least 10 ecstasy pills (MDMA users). This cut-off has also been used by other authors [26-28]. Subjects who used MDMA one or more times but less than the aforementioned amounts were not entered into the statistical analyses to ensure correctly separated groups. MDMA users and non-users were compared with regard to the following possible confounding variables by means of independent samples t-tests: age, general intelligence (Raven score), number of days since last cannabis use, duration of regular cannabis use before the initial assessment and duration of regular cannabis use between the first and the second assessment. Cannabis use was defined as duration of regular use because this measure, compared to age of first use, the average frequency of use, the cumulative life-time dose and the average dose per occasion, has been suggested to have the greatest impact on cognitive performance [29]. In order to operationalize health behaviour we performed a factor analysis (extraction method: principal component analysis; rotation method: varimax with Kaiser normalization) with all scores of the health behaviour questionnaire. The computed factor scores were also considered as potential confounders. Analogously, MDMA users and non-users were compared with regard to these variables by means of independent-samples t-tests.

Change scores of all cognitive variables were computed by subtracting the scores of the follow-up assessment from the baseline scores. Thereupon, we conducted three multivariate analyses of variance (MANOVA) with MDMA use (0 versus ≥10 pills) as fixed factor. The first MANOVA addressed attention and information processing speed (Trail making test part A, Stroop task parts A and B, digit symbol test). The second MANOVA addressed episodic memory (AVLT indices, LGT 3 indices). The third MANOVA addressed frontal/executive functioning indices (Trail making test part B, Stroop task part C, digit span test backwards). Moreover, we computed the effect size for each significant difference, operationally defined as η². All analyses were performed with IBM SPSS statistical software program version 19 (Chicago, IL, USA).

Results

Health behaviour

Concerning the health behaviour questionnaire, the factor analysis (extraction method: principal component analysis; rotation method: varimax with Kaiser normalization) revealed six factors with an eigenvalue greater than 4. The explained percentage of variance, the numbers of items of each factor and the corresponding eigenvalue, as well as their thematic attribution, are given in Table 1. For each factor the corresponding score per subject was output as a new variable. MDMA users and non-users did not differ with regard to any of these variables. Therefore, it was not reasonable to include these variables in the following MANOVAs as covariate. Corresponding means, SDs and significance levels are also given in Table 1. For example, the variables that belong to the factor associated with subjective wellbeing consist of items such as ‘On the following scale, please indicate how you rate your temporary wellbeing’ (‘… how satisfied you are with your …’ life/job/partnership/leisure time, etc.).

Table 1. Factor analysis of the health behaviour questionnaire.

	Factor 1 (sleep)	Factor 2 (alcohol use)	Factor 3 (sport/nutrition)	Factor 4 (subjective wellbeing)	Factor 5 (medical treatment)	Factor 6 (cigarette use)
Explained percentage of variance, number of items, eigenvalues, and mean factor values for each users group relating to the factor analysis of the health behaviour questionnaire. Standard deviations are given in parenthesis (extraction method: principal component analysis; rotation method: varimax with Kaiser normalization). ^aComputed by means of independent-samples t-tests between mean factor values of non-users and MDMA users (d.f. = 64).
Variance explained	7.36%	5.52%	5.21%	5.12%	4.67%	4.45%
Number of items	13	14	10	10	9	8
Eigenvalue	9.13	6.94	6.46	6.35	5.80	5.39
Non-users	−0.18 (±0.92)	0.08 (±0.93)	−0.06 (±1.07)	−0.14 (±0.87)	−0.03 (±0.97)	0.05 (±0.90)
MDMA users	0.14 (±1.20)	−0.09 (±1.18)	0.25 (±0.73)	−0.14 (±1.19)	0.02 (±0.79)	−0.06 (±0.95)
T-/P-valuea	−1.20/0.234	0.64/0.526	−1.26/0.165	0.02/0.983	−0.19/0.849	0.03/0.978

Group characteristics

Of the initial 149 subjects who participated in the first assessment, 109 subjects participated in the second assessment after 1 year. Forty-three subjects did not use any other illicit substance apart from cannabis over the course of the 1-year period (non-users). Twenty-three subjects used more than 10 pills MDMA (MDMA users; mean = 33.6; SD = 7.2; range: 10–62; mean occasions: 13.5; SD = 10.1; range: 4–36). The remaining subjects used MDMA once or more but less than the aforementioned amount and were not included in the further analyses (see the paragraph on statistical analyses). For each group, gender distribution, mean age, mean years of education, mean duration of cannabis use before the first assessment, mean duration of cannabis use between the first and the second assessments, mean days since last cannabis use at the second assessment, RAVEN score and corresponding SDs are given in Table 2. The groups did not differ significantly with respect to any of the variables. Therefore, it was not reasonable to include any of the aforementioned covariates in the following analyses. Corresponding significance levels are also given in Table 2. Moreover, there was no considerable use of cocaine, hallucinogens, sedatives, solvents/inhalants or opioids in the sample. In addition, the groups (user, non-user) did not differ with regard to their use of these substances. Therefore, it was not reasonable to enter these variables as covariates in the multivariate analysis. As usual in MDMA users’ samples, there was a high concomitant use of amphetamine (correlation MDMA and amphetamine use: r = 0.571, P = 0.004). Corresponding means and standard deviations of concomitant substance use are given in Table 3. Urine screens of all participants that were included in the analyses were free of amphetamines, benzodiazepines, cocaine, methadone and MDMA. Additionally, hair samples taken randomly by the Institute of Legal Medicine of the University of Cologne confirmed the self-reported substance use in all cases but one (which was excluded from the analyses).

Table 2. Group characteristics.

	Female/male	Age	Cannabis use at baseline ^a	Cannabis use within follow-up ^a	Days since last cannabis use ^b	Raven score ^c
Frequency of gender, mean age, mean duration of cannabis use, mean days since last cannabis use and mean Raven score for each users group. Standard deviations are given in parenthesis. ^aRegular use measured in months. ^bMeasured at second assessment. ^cMeasure of general intelligence (lower scores indicate better performance). ^dComputed by means of independent-samples t-tests between non-users and 3,4-methylenedioxymethamphetamine-users (MDMA) users (d.f. = 64). ^eComputed by means of χ^2 test.
Non-users	15/28	23.45 (±4.30)	41.52 (±5.24)	5.39 (±0.87)	267.81 (±101.47)	6.59 (±1.07)
MDMA users	09/14	25.52 (±6.51)	44.96 (±12.66)	4.74 (±1.17)	311.26 (±131.51)	9.09 (±1.43)
T-/P-value ^d	0.117/733 ^e	−1.44/0.155	−0.29/0.770	0.45/0.654	−0.26/0.798	−0.94/0.351

Table 3. Concomitant illicit substance use.

	Cocaine ^a	Hallucinogens ^b	Sedatives ^b	Solvents/inhalants ^b	Amphetamine ^a	Opioids ^b
Mean cocaine, hallucinogens, sedatives, solvents/inhalants, amphetamine and opioids use for each users group between the first and the second assessment. Standard deviations are given in parenthesis. ^aCumulative use between both assessments measured in grams. ^bCumulative use between both assessments measured in occasions. ^cComputed by means of independent-samples t-tests between non-users and 3,4-methylenedioxymethamphetamine-users (MDMA) users (d.f. = 64).
Non-users	0.52 (±1.55)	0.06 (±0.25)	0.06 (±0.25)	0.13 (±0.34)	2.9 (±5.40)	0 (±0)
MDMA users	1.04 (±2.42)	0.22 (±1.04)	0 (±0)	0.04 (±0.21)	26.55 (±33.91)	0.13 (±0.34)
T-/P-value ^c	−0.43/0.671	−0.55/0.586	1.06/0.295	0.75/0.941	−3.46/0.002	−1.82/0.083

Performance effects

The MANOVA addressing attention and information processing speed revealed no significant main effect of group (non-users versus MDMA users) (F _(4,61) = 2.06, P = 0.097). Mean test scores for both groups as well as significance levels of the corresponding tests of between-subjects effects are given in Table 4. The MANOVA addressing episodic memory revealed a significant main effect of group (F _(8,57) = 2.23, P = 0.043). The corresponding tests of between-subjects effects revealed significant effects of immediate recall (F _(1,64) = 11.43, P = 0.001, η² = 0.136) and delayed recall (F _(1,64) = 11.08, P = 0.002, η² = 0.144) of the LGT 3, but not for any AVLT variable. Mean test scores for MDMA users and non-users as well as significance levels of the corresponding tests of between-subjects effects are given in Table 5. Additionally, for each group, means and standard deviations of both LGT 3 variables are given in Fig. 1. The MANOVA addressing frontal/executive functioning indices revealed no significant main effect of group (F _(3,62) = 2.69, P = 0.054). Mean neuropsychological test scores for MDMA users and non-users as well as significance levels of the corresponding tests of between-subjects effects are given in Table 6. When entering amphetamine use as a covariate into the multivariate analysis, the corresponding effects did not remain significant. However, an analogous multivariate analysis with amphetamine use as grouping variable (0 versus 5 g) revealed no significant effects.

Figure 1. Means and standard deviations of both immediate and delayed recall Lern- und Gedächtnistest (LGT) 3 change scores for non-users and 3,4-methylenedioxymethamphetamine (MDMA)users

Table 4. Mean neuropsychological test scores for 3,4-methylenedioxymethamphetamine (MDMA) users and non-users related to attention and information processing speed

Test	MDMA users			Non-users			Sig.*	Hedge's G
	Mean (SD) (n = 23)			Mean (SD) (n = 43)
	Baseline	Follow-up	Change score	Baseline	Follow-up	Change score
*P-values refer to tests of between-subjects effects of the corresponding multivariate analyses of variance (MANOVA). ^aStroop task A: reading condition. ^bStroop task B: colour naming condition. SD: standard deviation.
Trail making test A	26.8 (7.2)	24.0 (10.4)	−2.87 (9.5)	25.4 (7.5)	22.3 (6.3)	−3.06 (8.6)	0.935	0.021
Stroop task A ^a	28.1 (4.2)	27.6 (8.8)	−0.56 (7.2)	29.3 (6.6)	27.2 (4.1)	−2.11 (6.9)	0.398	0.221
Stroop task B ^b	46.4 (9.5)	43.5 (7.98)	−2.90 (5.5)	44.5 (7.5)	42.4 (6.6)	−2.00 (4.2)	0.465	−0.192
Digit symbol test	60.8 (9.9)	63.8 (10.2)	3.00 (4.6)	64.3 (10.4)	69.0 (10.3)	4.67 (5.0)	0.189	−0.343

Table 5. Mean neuropsychological test scores for 3,4-methylenedioxymethamphetamine (MDMA) users and non-users related to episodic memory.

Test	MDMA users			Non-users			Sig.*	Hedge's G
	Mean (SD) (n = 23)			Mean (SD) (n = 43)
	Baseline	Follow-up	Change score	Baseline	Follow-up	Change score
*P-values refer to tests of between-subjects effects of the corresponding multivariate analyses of variance (MANOVA). Significant P-values are given in bold. ^a Rey Auditory Verbal Learning Test (RAVLT) A: immediate recall. ^bRAVLT B: total acquisition. ^cRAVLT C: recall after interference. ^dRAVLT D: loss after interference. ^eRAVLT E: recognition performance. ^fRAVLT F: repetitions required for learning. ^g Lern- und Gedächtnistest (LGT) 3 A: immediate recall. ^hLGT 3 B: delayed recall. SD: standard deviation.
RAVLT A ^a	7.17 (2.3)	6.91 (1.7)	−0.26 (2.9)	7.49 (1.9)	7.40 (2.2)	−0.09 (2.4)	0.919	−0.066
RAVLT B ^b	52.9 (9.8)	53.3 (8.4)	0.35 (9.6)	56.0 (8.0)	56.6 (8.2)	0.61 (6.5)	0.553	−0.034
RAVLT C ^c	11.4 (2.1)	11.1 (2.1)	−0.31 (2.1)	12.0 (2.5)	12.3 (2.9)	0.30 (2.7)	0.438	−0.243
RAVLT D ^d	1.48 (1.7)	1.57 (2.4)	0.09 (2.1)	1.65 (1.7)	1.51 (1.8)	−0.14 (2.6)	0.789	−0.094
RAVLT E ^e	13.9 (1.5)	13.7 (1.2)	−0.17 (1.3)	14.0 (1.8)	14.1 (1.5)	0.05 (2.0)	0.378	−0.123
RAVLT F ^f	4.65 (0.7)	4.26 (1.2)	−0.39 (1.1)	4.21 (1.1)	4.14 (1.1)	−0.07 (1.2)	0.280	−0.274
LGT 3 logos A ^g	12.4 (3.9)	10.7 (2.9)	−1.78 (3.5)	12.3 (3.6)	12.6 (2.9)	0.26 (2.8)	0.001	−0.667
LGT 3 logos B ^h	11.8 (4.0)	10.4 (3.3)	−1.35 (3.4)	11.2 (3.9)	12.3 (3.2)	1.09 (3.2)	0.002	−0.746

Table 6. Mean neuropsychological test scores for 3,4-methylenedioxymethamphetamine (MDMA) users and non-users related to frontal/executive functioning indices.

Test	MDMA users			Non-users			Sig.*	Hedge's G
	Mean (SD) (n = 23)			Mean (SD) (n = 43)
	Baseline	Follow-up	Change score	Baseline	Follow-up	Change score
*P-values refer to tests of between-subjects effects of the corresponding multivariate analyses of variance (MANOVA). ^aStroop task C: interference condition. SD: standard deviation.
Trail making test B	69.4 (22.2)	57.8 (25.9)	−11.6 (19.5)	63.1 (21.0)	50.8 (18.3)	−12.3 (18.4)	0.881	0.037
Stroop task C ^a	78.3 (16.1)	67.8 (18.6)	−10.5 (14.2)	72.1 (13.0)	67.7 (11.7)	−4.46 (10.0)	0.051	−0.520
Digit-span backwards	7.65 (2.3)	8.57 (2.6)	0.91 (1.93)	8.02 (1.9)	8.02 (2.3)	0.00 (2.3)	0.107	0.417

Discussion

The aim of the present investigation was to examine the nature of cognitive deficits in new ecstasy users over the course of a 1-year-period. A neuropsychological test battery including tests of learning, memory, working memory and executive functions was administered to 149 subjects, of whom 109 participated in the second assessment after 1 year. In addition, there were no statistically significant differences on a comprehensive number of possibly relevant confounders including age, general intelligence, cannabis use, alcohol use, cigarette use, medical treatment, participation in sports, nutrition, sleep patterns and subjective wellbeing. Significant effects of immediate and delayed recall of a paired associates learning task between subjects who used 10 or more ecstasy pills and subjects who did not use any illicit substance apart from cannabis during the course of the year (non-users) were found. No significant differences were found on any of the other neuropsychological tests.

Deficits in visual and associative learning among ecstasy users have been described previously [9, 30-33]. Due to the cross-sectional design of available studies, it was not possible to determine whether the found alterations existed before initiation of use or whether concomitant health behaviour variables were responsible for a percentage of the deficits. In addition, the role of concomitant use of cannabis in this context was not fully understood. In the present study, these methodological concerns were dispelled and a comparably large sample was investigated. Most intriguingly, although pre-existing group differences were ruled out and a comprehensive number of possible confounders were controlled for, the effects of ecstasy on paired associates learning remained significant despite the relatively short time-period (1 year) and the amounts of MDMA used (10–60 pills MDMA, mean: 32.44). Keeping in mind that no significant effect of MDMA on any other cognitive variable was found in the present study, the results indicate a specific effect of relative small amounts of MDMA on paired associates learning. This finding is consistent with the findings of Brown and colleagues [34], who also found significant deficits on an associate learning task in the absence of deficits on other memory tasks. Whether or not the influence of MDMA on other cognitive variables becomes significant with increasing time-periods and amount of use remains unclear, and is a promising hypothesis for future longitudinal investigations. In this respect, our results are inconsistent with the conclusions made by Halpern and colleagues [35]. They suggested that their findings ‘might instead reflect correctly that illicit ecstasy use, by itself, does not generally produce lasting residual neurotoxicity’. Potential reasons for the discrepant results could be found in the nature of the samples and the choice of measures that were administered. Because these issues have been already discussed extensively, we refer to the corresponding publications [36-38].

With regard to the intercorrelation between MDMA and amphetamine use in our MDMA users sample, it is unclear if the decrements in visual relational memory can be ascribed to the use of MDMA alone or to the polydrug use of MDMA and amphetamine. However, Gouzoulis-Mayfrank and colleagues [9] reported a significant influence of MDMA use on the immediate and delayed recall score in the same test that was used in the present study. Interestingly, in their sample it was possible to ascribe the effects to MDMA use by means of linear stepwise regression analysis. Therefore, it seems most likely that the decrements in visual relational memory can be ascribed to the use of MDMA rather than to the polydrug use of MDMA and amphetamine. Furthermore, after taking into account the potential effect of the use of other illicit drugs by means of multiple regression analyses, Schilt and colleagues [27] reported that ecstasy use was dose-related to verbal memory impairments, and there was still a significant association thereof.

Given that the hippocampus plays a fundamental role in relational memory [39], the findings of the present study support the hypothesis that the neural basis for the detrimental effects of MDMA on neurocognition appears to be a hippocampal dysfunction. Studies using animal models have shown that MDMA causes selective and persistent lesions of central serotonergic nerve terminals [4, 5, 40-42]. In addition, Kish and colleagues reported serotonin transporter changes in human MDMA users [43]. In particular, the hippocampus and parahippocampus display relatively high rates of serotonergic denervation after MDMA exposure [4]. Moreover, serotonin plays a central role in hippocampal neurogenesis inherent to learning and memory processes [44]. As a result, Gouzoulis-Mayfrank and colleagues [9] proposed that this may explain why hippocampal systems might be more sensitive to serotonin depletion than neocortical brain regions. Furthermore, it possibly accounts for the pattern of cognitive performance in ecstasy users with more consistent decrements in memory and learning performance compared to other cognitive domains that are dependent only on neocortical function. As mentioned previously, Brown and colleagues [34] also found significant deficits on an associate learning task in the absence of deficits on other memory tasks. However, they also found no deficits in ecstasy users on tests which have been shown to be more specific to the hippocampus and therefore interpreted their findings to complex interactions between multiple brain regions including the prefrontal cortex rather than to the hippocampus alone. The contributory role of other brain areas has also been illustrated by Burgess and colleagues who found reduced left parietal lobe activity during word recognition in abstinent MDMA users [45].

The underlying molecular mechanisms of these long-term effects have yet to be elucidated. Studies in laboratory animals have supported the involvement of oxidative stress in MDMA neurotoxicity by decreasing the levels of antioxidants in serotonergic terminals [46, 47]. Additionally, mitochondrial dysfunction and hyperthermia have been associated with administration of MDMA and subsequent toxicity to serotonergic terminals. Moreover, inflammatory cytokines, the ubiquitin proteasome system, environmental stress, neurotrophic factors and apoptotic proteins have been reported recently as potential mediators of MDMA toxicity that may also explain the terminal as the somatic degeneration [47]. In addition, by affecting basal synaptic transmission and long-term potentiation (LTP; an activity-induced increase in synaptic efficacy) via activation of serotonin receptors systems in the hippocampus, decreases in memory performance seem likely. Deciphering these molecular mechanisms is a promising task for further research. Moreover, neuroregulatory mechanisms are suggested to underlie MDMA-induced serotonergic dysfunction rather than neurodegeneration [11]. Similarly, this issue is an important target for future studies.

There are some methodological limitations which are inherent to open-trial studies and are outlined below. First of all, although we chose a prospective approach, the design was not experimental and therefore a causal relationship between MDMA use and decline in paired associates learning may not be presumed as a matter of course. Additionally, for reasons of practicability the minimal abstinence period from cannabis in the present study was 12 hours. Therefore, we cannot rule out an impure distinction between subacute and long-term effects of cannabis use. However, the mean time since last cannabis use did not differ between MDMA users and controls. Furthermore, the quantity of use was reported by the participants themselves. Nevertheless, studies validating self-reported voluntary substance use found a high reliability of the reported drug quantity [48-50]. Additionally, hair samples taken randomly by the Institute of Legal Medicine of the University of Cologne confirmed the self-reported substance use in all cases but one (which was excluded from the analyses). Another limitation concerns the concentration and purity of illicit drugs and their mode of consumption. However, evidence from German police seizures suggest that 99.65% of the confiscated ecstasy pills in 2008 contained only one psychoactive ingredient. MDMA was found in 96.8% of the confiscated ecstasy pills. The remaining share of 3.2% of the confiscated ecstasy pills consisted of 1-(3-chlorphenyl)-piperazine (m-CPP), amphetamines, methamphetamine or MDA [51]. Cannabis is a natural product, inherently variable in its strength and composition. Therefore, estimates of the life-time cumulative exposure of illicit drugs represent only crude approximations of actual exposure [3, 52]. Furthermore, subjects were aware that their drug use was critical to the research design, and this awareness may have caused selection bias or created expectation effects, and subjects who took ecstasy may have experienced anxiety about confirming [53]. However, the significant differences were found specifically in paired associates learning, and not in other domains of cognitive functioning.

In conclusion, a significant effect of MDMA use on visual paired associates learning was found, suggesting specific deterioration of hippocampal functioning. Given the fact that memory impairments remained significant after controlling for a large variety of confounders, our findings may raise concerns with regard to MDMA use, even in recreational amounts over a relatively short time-period.

Declarations of interest

This work was supported by a grant to E. Gouzoulis-Mayfrank and J. Daumann from the German Research Foundation (Deutsche Forschungsgemeinschaft DFG, Go 717/6-1/2). Furthermore, the authors declare that, except for income received from their primary employer, no financial support or compensation has been received from any individual or corporate entity over the past 36 months for research or professional service and there are no personal financial holdings that could be perceived as constituting a potential conflict of interest. This pertains to all the authors of the study, their spouses or partners and their children (aged under 18).