By necessity, the discussion of methodological constraints had to be very concise in the published version of our paper (it was more detailed and included some of the arguments raised by Heinz & Jones in the originally submitted manuscript). We are therefore glad to have this opportunity to respond to the constructive comments of Heinz & Jones. They essentially make two claims: that β-CIT does not reliably label cortical serotonin transporters, so that our observed group difference must be due to an alternative mechanism; and at 90 minutes after tracer injection there is a significant admixture of other effects, such as blood flow, blood-brain barrier integrity and tissue permeability, with the same result.

The first claim is supported by some, but not all, displacement studies with SSRIs in monkeys, but inter-species comparisons of brain measures have to be judged with reserve, as Heinz & Jones point out. They also cite a very recently published abstract of a study in six humans, using an alternative (PET-) ligand. We look forward to the full paper; if the initially reported claim survives peer review, it may certainly call into question our interpretation. Moreover, it will specifically weaken the McCann et al (Reference McCann, Szabo and Scheffel1998) paper, whose authors used the same PET tracer and found cortical reductions in serotonin transporter labelling. The design of our study was based on Kuikka et al's (Reference Kuikka, Tijhonen and Bergström1995) original report. They examined relatively large numbers of healthy volunteers (28) and patients (9) at one and two hours after injection of β-CIT. They reported significant tracer washout with 20 mg citalopram from medial prefrontal cortex (Brodmann's area 12) in 25 subjects, 1-2 hours after injection. They also found significant specific binding of serotonin transporters in occipital cortex. Both are regions that showed activity reductions in our MDMA users. They further described reduced medial prefrontal cortex β-CIT activity at 1 hour in five (alcoholic) patients compared with controls, in the absence of perfusion differences measured with the single photon emission computed tomography (SPECT) ligand ^99mTc-ethyl cysteinate dimer.

The second (weaker) claim made by Heinz & Jones is correct in the sense that group differences in β-CIT binding at 90 minutes do not necessarily reflect a difference in serotonin transporter binding. However, in the absence of a priori hypotheses about generalised cell loss, reductions in blood flow, increased blood-brain barrier integrity or reduced tissue permeability, our results are at least consistent with our interpretation. We think the writers over-state their point if they claim that at one hour non-specific factors are ‘prevalent’ in determining binding (Kuikka, 1995). Our β-CIT images clearly show activity patterns that parallel the known distribution of serotonin transporters, with relatively high activity in midbrain (Fig. 1 in Semple et al, 1999).

An important experiment that has not yet been performed is the displacement of β-CIT binding by ‘cold’ serotonin transporter ligands (e.g. citalopram) in areas that are found to be abnormal. It needs to be emphasised, however, that the more specific investigations also tend to be more invasive (e.g. PET with arterial blood sampling) or more of a burden to the subject (e.g. dynamic SPECT scan 4-24 hours after tracer injection with citalopram, resulting in corresponding increases in radiation dose or scan time). This can potentially increase measurement error and aggravate the selection bias of the study, thereby reducing its validity. What is gained in theoretical experimental power may well be lost in spurious or biased sampling, if subjects have to be paid to participate (ours were not) or if subjects are self-selected on the basis of some perceived problem. It behoves the reader to be sceptical about any claims based on small samples, as well as non-specific methodologies, and to scan the medical literature for replicable results, keeping in mind that there is publication bias in favour of positive findings. As far as MDMA-induced damage to human serotonin neurons is concerned, the jury is clearly still out.