Statistics
Power analysis
Single photon emission connputed tomography, 'HMRS,
DTI, PWI, and memory performance
Previous SPECT findings (Reneman et al., 2001a)
indicated that 8 females (effect size d= 0.16) and 31
males (d=0.08), thus a group of 39 subjects per group,
would provide ample power to demonstrate a difference
in SERT densities between pre-post assessments
(prospective study) and between lifetime ecstasy users
and ecstasy-naive controls (retrospective studies), if
such a difference exists. This power estimate is in general
agreement with two other imaging studies
conducted with ecstasy users, which showed that 25
subjects were needed per group to demonstrate a difference
in serotonergic transporter densities using PET
(males and females together) (McCann et al., 1998)
and 32 using SPECT (only males) (Semple et al.,
1999). The sample sizes in all three substudies would
also be big enough to detect effect sizes of 28%, 21%,
and 31% on outcome parameters measured with 'HMRS,
PWI, and memory performance (especially on
RALVT) respectively, as indicated by previous studies
(Reneman et al., 2001b; Reneman et al., 2001c;
Reneman et al., 2002b).
Eunctional MRI
Previous studies indicated that changes in cognitive
abilities are small but significant after moderate ecstasy
use. For reliable measurement of cognition-related
functional brain activity patterns a sample size of
about 10 to 12 subjects is required. To detect differences
reliably between ecstasy users and controls,
10-12 subjects would be required per group. As brain
activity patterns might differ between male and female
ecstasy users, 20 to 24 subjects per group would be
required in order to obtain representative samples for
both genders.
Statistical analyses
We hypothesized a priori that if ecstasy use is indeed
neurotoxic, ecstasy users would differ on various imaging
parameters (for example, increase of ADC, rCBV,
Cho, ml and decrease of ['^^I]|3-CIT uptake ratios, FA,
NAA), on BOLD fMRI parameters (increased activity
or alterations in patterns of activation), as well as on
parameters of neurocognitive functioning (such as
decreased memory) and psychopathology (such as
increased depression, impulsivity, sensation seeking)
compared with non-users (cross-sectional substudy and
retrospective cohort substudy) or compared with their
own baseline values prior to first ecstasy use (prospective
cohort substudy).
For the cross-sectional substudy among subjects
with variations in amount and type of drugs used, parameters
of neurotoxicity will be assessed using linear
multiple regression analysis with lifetime use of ecstasy.
Neurotoxicity of ecstasy
179 cannabis, amphetamine, and cocaine as separate
regressors. It is expected that this will provide information
about the relative contributions of the various
drugs on the main outcome parameters. The regression
model will also control for factors other than drug use,
such as gender, age, and DART-IQ.
For the prospective cohort substudy, follow-up data
will be compared between incident ecstasy users and
persistent ecstasy-naive subjects using (multivariate)
analysis of variance (ANOVA/MANOVA), including
baseline measurements and significant confounders
(such as age, gender, use of cannabis, amphetamines
and cocaine) as covariates (ANCOVA, MANCOVA).
In order to prevent the loss of subjects due to incomplete
data, general linear mixed models could be
applied in the analysis of the longitudinal data.
For the retrospective cohort study, parameters of
neurotoxicity will be compared cross-sectionally
between lifetime ecstasy users and matched non-users.
An analysis of covariance will be used with main confounders
(such as age, gender, cumulative dose of
ecstasy, use of cannabis, internalizing and externalizing
psycbopathology at age 4-16 measured with tbe
CBCL, prior to first ecstasy use in tbe group of lifetime
ecstasy users) as covariates. Correlations between
characteristics of ecstasy use (sucb as lifetime CD,
duration of abstinence) and outcome parameters will
be analysed using a linear regression analysis.