Moncrieff & Leo (2010) provide an excellent overview of the literature on the association between antipsychotic medication intake and global brain volume changes. The authors build the case that at least some of the brain abnormalities that are found in schizophrenia patients are not a consequence of the illness itself but result from antipsychotic medication. Indeed, being ill is in the majority of cases linked to the intake of antipsychotic medication, which is known to interact with the brain. The underlying mechanisms of how the structure of the brain is influenced by antipsychotic medication are largely unknown. Trying to understand these mechanisms is important in order to put the findings from structural magnetic resonance imaging (MRI) studies into perspective.
Two issues are important to take into account when interpreting the longitudinal MRI studies that include medicated patients. First, as antipsychotic medication interacts with neurotransmitter systems it is not unlikely that it affects the brain focally, as was also suggested by Navari & Dazzan (2009) in their overview of the literature on this topic. By investigating global brain volumes no conclusion can be drawn on focal abnormalities. Second, as different antipsychotic drugs act on different neurotransmitter systems it would not be surprising that type of medication is a crucial factor. Indeed, two of the largest follow-up MRI studies (including the only randomized trial) (Lieberman et al. 2005; van Haren et al. 2008) found evidence for differential effects from different kinds of antipsychotics on global brain volumes. This is of particular interest since it was suggested that typical antipsychotics were related to a more pronounced loss while atypical medication (olanzapine in particular) was related to less pronounced loss. This was not only the case in first-episode patients (Lieberman et al. 2005) but also in a sample consisting of first-episode and chronically ill patients who had all been medicated for more than a few weeks at baseline measurement (van Haren et al. 2008). Moreover, to investigate the effects of medication in an unbiased fashion randomized control trials are essential; therefore, it is justified to place more weight on the results of the only large randomized trial, even though it was funded by the manufacturers of olanzapine.
It is even more important to realize that not finding brain volume abnormalities in medication-naive patients does not necessarily indicate that the brain abnormalities occur as a result of starting antipsychotic medication, even though the timing might suggest this. Much of the argument that loss of brain tissue is related to intake of medication is built on this assumption. It might well be that the brain starts to change at illness onset, that it progresses during the course of the illness, and is actually a consequence of being ill. The medication-naive studies cannot provide evidence for either one of these hypothesis.
Many studies found strong evidence for brain volume abnormalities to be related to outcome. Those patients with a poorer outcome show more pronounced loss of brain tissue compared to those with a better outcome. This indicates that the illness process itself is at least associated (not to say responsible). Indeed, based on some of our own findings that the effects of outcome on brain volume appear to be independent of the intake of medication (Cahn et al. 2006; van Haren et al. 2008), it could be that progressive brain volume change is indeed related to both outcome and medication intake independently.
The authors suggest that future studies should randomize first-episode patients to either treatment with antipsychotic medication or to withhold antipsychotic treatment for a few weeks. Studies like this are extremely difficult to perform. It is easier, and not less informative, to perform randomized controlled trials like the one that has been done by Lieberman et al. (2005) and compare patients using different types of medication in a longitudinal fashion or study the effects of (randomized) discontinuation of medication.
Drs van Haren and Cahn make several interesting comments. The most interesting comment concerns the ‘assumptions’ of the neuroimaging field. It is these assumptions about medication which are the heart of the question, or disagreement, when it comes to interpreting the imaging data. Our paper called into question a strongly held assumption in the schizophrenia imaging field (Moncrieff & Leo, 2010), namely, that any observed difference between the brains of patients and controls can be attributed to an organic pathology. Yet studies using only medicated patients cannot provide evidence of an organic pathology, that is, unless one makes the prior assumption that the medications are not a confounding variable.
They point out in their letter the problem with interpreting results from patients who are medication naive at the point of initial scan, and show volume reductions following a course of antipsychotic medication. Are the observed reductions due to the natural course of the disease or to the medications? We agree that without well-designed trials the results of these studies are hard to interpret, which is why we suggested the need for more well-developed studies.
Data on clinical outcomes and brain-volume changes are also difficult to interpret, and may also reflect medication-induced effects, since people with worse outcomes may receive more medication. In the neuroimaging field indications of brain volume reduction over time were found to be associated with poor outcome in several of the longitudinal studies examined (Davis et al. 1998; Lieberman et al. 2001; Mathalon et al. 2001; Cahn et al. 2002; Ho et al. 2003; Nakamura et al. 2007), although others found no effect or opposite effects (Sporn et al. 2003; DeLisi et al. 2004). Other factors that may reflect exposure to antipsychotic treatment, such as longer duration of hospital admission, duration of illness and increased number of hospital admissions, were also associated with reduced brain volume in several of these studies (Mathalon et al. 2001; DeLisi et al. 2004; van Haren et al. 2008).
The study cited to support the argument that neuroimaging studies still provide evidence of an organic pathology, examined brain-volume changes over a 5-year period (Cahn et al. 2006). But in this study all of the patients received medication during the 5-year period between scans (confirmed by an email to one of the authors). Without a medication-naive group of patients it seems problematic to assume that the medications did not play a role in the observed volume reduction in a group of people exposed to antipsychotic medications for 5 years. No data on the association between antipsychotic exposure and brain volumes at 5-year follow-up has yet been published, but 1-year follow-up of this cohort showed that cumulative dose predicted loss of grey-matter volume (Cahn et al. 2002).