----- Forwarded message from Kevin Black <[log in to unmask]> -----
>gets. A different but related issue is that for most drugs, you cannot
>remove their effects quickly from the body.
In itself, this does not imply that fMRI is less appropriate than O-15
PET (i.e., both are sensitive to movement artifact). If anything, the
short half-life of 15-O, in combination with the limited number of times
it can be administered serially, would argue in favor of using fMRI.
>If a drug affects coupling of metabolism to blood flow, or if a
>drug has a global effect on both, how will you interpret a
>change in BOLD signal in a certain region?
With regard to drugs that modulate the coupling between neuronal (glial?)
metabolism and blood flow, I would imagine that this would impact O-15 PET
analyses, which are based upon the assumption that rCBF is monotonically
related to regional neuronal activity (mean multiunit firing rates). As an
aside, Goldman-Rakic's lab has recently published data suggesting that
dopamine can directly alter physiological paramters related to blood flow.
The effects of other neurotransmitters, such as acetylcholine or
norepinephrine, on these parameters -- at least in the periphery -- has
long been recognized.
With regard to drugs that affect gCBF, gCMR, or both, my assumption is
that this can be partially (?) corrected by removing the global signal
change prior to regressing the modelled treatment effects. This is
predicated upon the assumption that global and regional effects of the
drug are minimally collinear (cf. D'Esposito group's [formerly UPenn;
now Berkely, Columbia, and UPenn] work). Given that the global signal is
typically partitioned prior to testing treatment effects in PET analyses,
again, I do not see why this possibility implies an advantage for O-15 PET
relative to fMRI.
>For that matter, how will you know that BOLD signal has changed?
>Most cognitive studies include baseline periods or trials interspersed
>from beginning to end of a series of image acquisitions. You can't
>do this for most drugs.
Not necessarily; this would depend on whether there exist a priori
predictions about specific cognitive or affective tasks that would be
modulated by the drug, e.g., D1 receptor antagonists and a working
There exist other means of modeling the treatment effect. One could
collect self-report measures (cf. Breiter et al.; see below), peripheral
physiology, or explicitly model the plasma drug curve (cf. Stein et al).
Although I'm not aware of any study that has done so, presumably one could
design a parametric experiment in which one superimposed different doses
on a time-varying cognitive or affective task. Assuming that the drug
modulated performance, one could presumably use the dose-response function
to model the changes in regional MR signal.
Since the BOLD signal is nonquantitative
>and in general is not predictable over time, you have to be tricky
>and/or very careful to try to interpret changes as meaningful.
This statement would seem to call into question the results of nearly any
study employing fMRI. As an aside, this is an old problem in the
world of animal behavioral neuroscience: 2DG autoradiography, cFOS, and
HPLC-based in vivo transmitter assays are all (often) done without
reference to a true ratio measurement system.
>There are groups who feel they have overcome these problems.
>You should read their papers before you design the experiment.
>Try MEDLINE with MRI + BOLD + (cocaine or methylphenidate).
In addition, I'd look into the work that Elliott Stein and his colleagues
at the Medical College of Wisconsin (Milwaukee, WI, USA) have published in
which they utilized fMRI to examine regional BOLD changes in response to
nicotine and/or cocaine. Their analytic path (partially instantiated in
MCW's AFNI software package) is much more sophisticated than that used by,
for example, Breiter and colleagues in their seminal cocaine study.
Alexander J. Shackman
Laboratory for Affective Neuroscience
Department of Psychology
University of Wisconsin - Madison
1202 West Johnson Street
Madison, Wisconsin 53706
EMAIL: [log in to unmask]