Salisbury et al., in the May 2007 issue of Archives of General Psychiatry, demonstrate associated progressive reductions in mismatch negativity (MMN) amplitude and Heschl’s gyrus reduction in schizophrenia. These findings provide strong support for involvement of auditory cortex in the pathogenesis of schizophrenia, and demonstrate that pathological changes in the illness are not confined to specific brain regions, such as prefrontal cortex, that receive the preponderance of attention.

Further, the manuscript helps resolve an important current controversy in the MMN literature. Deficits in MMN generation have been among the most consistent findings in chronic schizophrenia, with a recent meta-analysis showing large (~1 sd unit) effect size MMN reductions across studies (Umbricht et al., 2005). As noted by Salisbury et al., however, deficits have not been observed in first-episode patients (Salisbury et al., 2002; Umbricht et al., 2006)....  Read more

Salisbury et al., in the May 2007 issue of Archives of General Psychiatry, demonstrate associated progressive reductions in mismatch negativity (MMN) amplitude and Heschl’s gyrus reduction in schizophrenia. These findings provide strong support for involvement of auditory cortex in the pathogenesis of schizophrenia, and demonstrate that pathological changes in the illness are not confined to specific brain regions, such as prefrontal cortex, that receive the preponderance of attention.

Further, the manuscript helps resolve an important current controversy in the MMN literature. Deficits in MMN generation have been among the most consistent findings in chronic schizophrenia, with a recent meta-analysis showing large (~1 sd unit) effect size MMN reductions across studies (Umbricht et al., 2005). As noted by Salisbury et al., however, deficits have not been observed in first-episode patients (Salisbury et al., 2002; Umbricht et al., 2006). An unknown issue was whether the discrepancy between first-episode and chronic patients was due to within-subject change (the “degeneration” hypothesis), or whether those patients with small MMN at entry tended to be retained disproportionately in chronic samples because of the relationship between MMN generation and global outcome (e.g., Light and Braff, 2005) (the “distillation” hypothesis).

The present study suggests that at least some patients show reductions of both MMN amplitude and left HG volumes over time, lending at least partial support for the degeneration hypothesis. This finding is important in that it shows that the pathological process contributing to cognitive impairment in schizophrenia continues beyond first episode, and may be a target for pro-cognitive interventions. It should be noted that the degeneration continued despite treatment with atypical, as well as typical, antipsychotic medication.

As noted by Salisbury et al., the change in MR volume in schizophrenia is best conceived as atrophy of neurons, rather than degeneration. On a histological level, the volume reductions noted on MR correspond with reduced pyramidal cell size in postmortem tissue (e.g., Sweet et al., 2004). Interestingly, postmortem studies have yet to show volumetric reductions in HG despite the change in some compartments, suggesting that MR may be detecting changes in tissue parameters that are not apparent in postmortem histological examination. This study also complements a recent diffusion tensor imaging (DTI) study that showed correlations between white matter changes in auditory projection pathways and auditory processing deficits in schizophrenia (Leitman et al., 2007). The relationship between white matter and grey matter pathology requires further investigation.

There are additional lessons hidden in the Salisbury et al. study. Given the relationship between reduced MMN generation (a functional measure) and cortical volume (a structural measure), there is a strong tendency to assume that structural changes are the cause of functional changes. The findings by Salisbury et al., as well as the extrapolation to postmortem histological studies, argue strongly against such an interpretation. For example, in the Salisbury et al. study, the change in left HG volume from time 1 to time 2 was only 6 percent, whereas MMN declined by 33 percent over the same period of time. At time 2, HG volumes were only 2 percent smaller in schizophrenia patients vs. controls, whereas MMN was 35 percent smaller. These findings suggest that simple volume loss does not cause the reduction in MMN. Further, even though MMN reduction seems to stabilize following the first 1.5 years (e.g., Umbricht et al., 2006; Javitt et al., 1995), this may not be the case with volumetric deficits. Thus, in a prior sample of chronic patients, this same group reported reductions of 13 percent in HG volume (Hirayasu et al., 2000), as opposed to the 2 percent reduction observed in patients following 1.5-year follow-up. Rather than suggesting a primary role of degeneration, this suggests a “use it or lose it” relationship within auditory cortex, wherein persistent reduction of activity may lead over time to structural involution. Even in postmortem studies (e.g., Sweet et al., 2004), pyramidal cell volumes are reduced by only 10 percent, whereas MMN in chronic schizophrenia may be reduced by 40 percent or more (e.g., Salisbury et al., 2002; Umbricht et al., 2006).

As noted by Salisbury et al., acute treatment with NMDA antagonists leads to reduced MMN amplitude in both human (Umbricht et al., 2000) and animal (Javitt et al., 1996) models. NMDA receptors also play a critical role in synaptic spine development and maintenance (Matsuzaki et al., 2004). A possible explanation, therefore, is that reduced NMDA activity in auditory cortex leads to both MMN reductions and reductions in spine density. Alternatively, primary alteration in subpopulations of cortical glutamatergic cells could trigger the sequence of events leading to reduced MMN generation.

There are several other intriguing features to the dataset. For example, at baseline, there were several controls who had larger than median HG volumes, but nevertheless failed to generate MMN (i.e., <1 μV). In schizophrenia patients, this sector of the plot was entirely empty and the only subjects who failed to generate MMN were those with small HG volumes. This suggests that there may be fundamental differences in structure/function relationships. It is almost as interesting to know why some controls fail to generate MMN despite having adequate HG size, as it is to know why HG is reduced in schizophrenia.

The finding that the relationships hold only for left, not right, HG, also is worthy of further investigation, as is the finding that right HG volumes are reduced even at first episode and do not decline further. Finally, the correlation on reduced MMN amplitude at Fz with reduced HG volume reiterates once again the role of auditory, rather than frontal, cortices in mediating MMN generation deficits in schizophrenia.

The authors reported a cross-sectional (first hospitalization or within 1 year of first hospitalization) and longitudinal (1.5-year follow-up) study of electrophysiologic testing (mismatch negativity, or MMN, amplitude) and high-resolution structural magnetic resonance imaging of Heschl gyrus and planum temporale gray matter volumes. Schizophrenia subjects showed longitudinal volume reduction of left hemisphere Heschl gyrus (P = .003), which was highly correlated with MMN reduction (r = 0.6; P = .04). The interrelated progressive reduction of functional and structural measures suggests progressive pathologic processes early in schizophrenia. The design of the study helped minimize the effect of medication, the authors commented, therefore allowing the interpretation that brain change is due to disease progression.

From an imaging perspective, this is a straightforward longitudinal study of brain structure following previously published image processing and measuring protocols (Kasai et al., 2003). T1- and T2-weighted MR scans...  Read more

The authors reported a cross-sectional (first hospitalization or within 1 year of first hospitalization) and longitudinal (1.5-year follow-up) study of electrophysiologic testing (mismatch negativity, or MMN, amplitude) and high-resolution structural magnetic resonance imaging of Heschl gyrus and planum temporale gray matter volumes. Schizophrenia subjects showed longitudinal volume reduction of left hemisphere Heschl gyrus (P = .003), which was highly correlated with MMN reduction (r = 0.6; P = .04). The interrelated progressive reduction of functional and structural measures suggests progressive pathologic processes early in schizophrenia. The design of the study helped minimize the effect of medication, the authors commented, therefore allowing the interpretation that brain change is due to disease progression.

From an imaging perspective, this is a straightforward longitudinal study of brain structure following previously published image processing and measuring protocols (Kasai et al., 2003). T1- and T2-weighted MR scans were acquired using the same sequence and on the same scanner for all subjects and at all time points. All baseline and follow-up MR scans were bias-field corrected and used in a fully automated segmentation algorithm for tissue classification, and then realigned to standard coordinate space and re-sampled to isotropic voxel resolution for application of standard manual segmentation protocols. Intracranial content was also estimated. Inter-rater and intra-rater reliability for segmentation of the Heschl gyrus and planum temporale was very high (volume ICC ranging from 0.95 to 0.99) (Kasai et al., 2003).

The authors showed in their earlier paper (Kasai et al., 2003) that using this approach, the time-dependent change in the volume of intracranial content did not correlate with time-dependent volume changes of brain structures. While this is reassuring, a trend-level decrease of intracranial content in time (p = 0.065), however, does raise the possibility of some systematic bias such as scanner drift resulting in global scaling, especially considering the subjects’ ages of 21-24 years. Some solutions such as scaling the follow-up scans with respect to the baseline scans could be evaluated (Freeborough and Fox, 1997).

This well-designed and well-presented study adds to a growing body of evidence that longitudinal structural neuroimaging is an effective way to detect progressive changes in specific brain structure in patients with schizophrenia. The results of this study contribute to the debate over whether the pathogenesis of schizophrenia includes a neurodegenerative as well as neurodevelopmental component.

Longitudinal increases in volume of the lateral ventricles and decreases in brain volume—progressive changes—are often observed over time early in the course of schizophrenia. There is not uniform agreement over the proper interpretation of these changes, prompting vigorous, healthy debate among investigators. A major point of contention appears to be whether these volume changes actually constitute evidence of active disease progression.

In the current study, the authors seek to bolster the case for structural progression by demonstrating evidence of interrelated progressive functional impairment. They buttress the case for structural progression by demonstrating a relationship between worsening deficit in mismatch negativity and auditory cortex volume decreases.

Identification of a direct causal relationship between the underlying pathophysiology of schizophrenia and volume losses observed early in the illness would conclusively demonstrate structural progression. Such a direct link has not yet been established, so the results of this study constitute...  Read more

Longitudinal increases in volume of the lateral ventricles and decreases in brain volume—progressive changes—are often observed over time early in the course of schizophrenia. There is not uniform agreement over the proper interpretation of these changes, prompting vigorous, healthy debate among investigators. A major point of contention appears to be whether these volume changes actually constitute evidence of active disease progression.

In the current study, the authors seek to bolster the case for structural progression by demonstrating evidence of interrelated progressive functional impairment. They buttress the case for structural progression by demonstrating a relationship between worsening deficit in mismatch negativity and auditory cortex volume decreases.

Identification of a direct causal relationship between the underlying pathophysiology of schizophrenia and volume losses observed early in the illness would conclusively demonstrate structural progression. Such a direct link has not yet been established, so the results of this study constitute only indirect evidence that structural progression is tied to the emergence of functional impairment. Results of longitudinal MRI studies are useful for identify factors potentially associated with these volume changes, including altered neurodevelopment, disease progression, mismatch negativity, antipsychotic medications, and yet unidentified factors. Until the underlying etiology of schizophrenia is known, what underlies longitudinal volume change in schizophrenia is unlikely to be determined.

Future research should focus on specifying the neurodevelopmental mechanisms that contribute to the cortical pathology central to schizophrenia.

We appreciate the SRF focus on our article (Price et al., 2005) and the comments by Robert Freedman, Danielle Dick, and other contributors to the general topic of endophenotypes. A couple of points raised call for a brief response.

Freedman’s query whether by combining several endophenotypes we implicitly assume that “overlapping sets of genes” are involved can be answered in the affirmative. It is now generally accepted that no 1:1 relationship exists between genes and phenotypes in the polygenic (or oligogenic) disorders. Similarly to the multiple interrelated neural systems, the sets of susceptibility and modifier genes operate as complex interacting networks that functional genomics is only now beginning to understand and tease out (see Liu et al., 2002; Jablensky, 2004). In this context, the requirement that the “genetic architecture”...  Read more

We appreciate the SRF focus on our article (Price et al., 2005) and the comments by Robert Freedman, Danielle Dick, and other contributors to the general topic of endophenotypes. A couple of points raised call for a brief response.

Freedman’s query whether by combining several endophenotypes we implicitly assume that “overlapping sets of genes” are involved can be answered in the affirmative. It is now generally accepted that no 1:1 relationship exists between genes and phenotypes in the polygenic (or oligogenic) disorders. Similarly to the multiple interrelated neural systems, the sets of susceptibility and modifier genes operate as complex interacting networks that functional genomics is only now beginning to understand and tease out (see Liu et al., 2002; Jablensky, 2004). In this context, the requirement that the “genetic architecture” of the relevant endophenotypes should be simpler than that of the clinical phenotype of schizophrenia appears to be unwarranted and need not be a defining criterion of an endophenotype (see Michael Owen’s contribution to the SRF endophenotype discussion). Further, the seeming paradox of little correlation among the individual endophenotypes and the increase in relative risk when they are combined is compatible with their variable individual expression being offset by an upstream latent trait (see Deborah Levy’s comment in the SRF endophenotype discussion ).

In our experience, the main advantage of multivariate or composite endophenotypes is twofold, that is, (a) allowing a substantial increase in effect size, relative risk and, consequently, power for genetic analysis; and (b) providing tools for reducing the notorious heterogeneity of schizophrenia by parsing the broad clinical phenotype into relatively homogeneous subtypes that may have a distinct genetic basis. In the Western Australian Family Study of Schizophrenia, our research group systematically phenotyped, over the last 8 years, 388 members of 112 families with one or more affected members and over 150 population controls for multiple neurocognitive, neurological, electrophysiological, and personality features. By analyzing the whole-genome scan of these families using a composite neurocognitive endophenotype, which integrates performance measures across several domains, we recently demonstrated (Hallmayer et al., 2005) that a distinct subset of schizophrenia families (including ~30 percent of the probands in our sample) shares a pervasive cognitive deficit linked precisely (lod score 3.32) to the locus on 6p24-22 previously reported by Straub and colleagues (Straub et al., 1995) in a large sample of Irish multiplex schizophrenia families. Further genetic analyses of the Western Australian cohort will aim to explore the potential contribution of the multivariate electrophysiological endophenotype to a refined subclassification of schizophrenia.

References:
Hallmayer JF, Kalaydjieva L, Badcock J, Dragovic M, Howell S, Michie PT, Rock D, Vile D, Williams R, Corder EH, Hollingsworth K, Jablensky A. Genetic evidence for a distinct subtype of schizophrenia characterized by pervasive cognitive deficit. Am J Hum Genet. 2005 Sep;77(3):468-76. Epub 2005 Jul 12. Abstract

Price GW, Michie PT, Johnston J, Innes-Brown H, Kent A, Clissa P, Jablensky A. A multivariate electrophysiological endophenotype, from a unitary cohort, shows greater research utility than any single feature in the Western Australian Family Study of Schizophrenia. Biol Psychiatry. 2005 Dec 17; [Epub ahead of print] Abstract

I appreciate Dr. Yung's comments on our pharmacotherapeutic treatment trial in a sample of young persons with "prodromal" symptoms and high risk for becoming psychotic within a short period of time. It was her work with Pat McGorry that first demonstrated this population could be identified, thus opening up the potential for prospective study of the mechanisms of onset and the study of treatment as preventive as opposed to merely ameliorative. We were concerned about the high dropout rate for obvious reasons, but in retrospect we should not have been surprised. Our sample was young and perhaps more resistant for that reason, as Dr. Yung implies, but the fact is that 2 years is a very long clinical trail no matter what the age! In part we wanted to allow sufficient time to elapse to capture higher numbers of converting subjects, and that still seems to be a reasonable strategy insofar as the conversion rate in the placebo group had not clearly plateaued by 1 year. Nevertheless, in retrospect, a trial of 2 years was unrealistic.

The optimal design, clearly, would be larger...  Read more

I appreciate Dr. Yung's comments on our pharmacotherapeutic treatment trial in a sample of young persons with "prodromal" symptoms and high risk for becoming psychotic within a short period of time. It was her work with Pat McGorry that first demonstrated this population could be identified, thus opening up the potential for prospective study of the mechanisms of onset and the study of treatment as preventive as opposed to merely ameliorative. We were concerned about the high dropout rate for obvious reasons, but in retrospect we should not have been surprised. Our sample was young and perhaps more resistant for that reason, as Dr. Yung implies, but the fact is that 2 years is a very long clinical trail no matter what the age! In part we wanted to allow sufficient time to elapse to capture higher numbers of converting subjects, and that still seems to be a reasonable strategy insofar as the conversion rate in the placebo group had not clearly plateaued by 1 year. Nevertheless, in retrospect, a trial of 2 years was unrealistic.

The optimal design, clearly, would be larger samples treated for a shorter period, but recruiting larger samples proved to be very difficult, even with four sites, which gets to Dr. Yung's final point. The true positive prodromal person/patient emerges in the population at the incidence rate of schizophrenia which is not robust (one in 10,000 per year). Furthermore, the earliest symptoms are often negative in nature and hard to identify, especially if the person is no longer living at home with family, that is, with people who might be sensitive to nuance changes. The bottom line is that research in this field is an uphill battle vis-à-vis sampling and recruitment. Yet I feel strongly that such samples are extraordinarily valuable for studies of the pathophysiology and preventive treatment of schizophrenia because, unlike retrospective studies, predictions that are prospectively falsifiable can be made and tested.

Therefore, like Dr. Yung, I endorse current efforts to consolidate samples across sites and to plan studies that are multisite so that sufficient numbers of such potentially informative patients can be gathered and pooled. In North America, for example, eight sites have used a common prodromal assessment battery (the same assessment instruments used in the Lilly clinical trial) and have pooled their data with the help of supplemental funds from NIMH. This group, called the North American Prodromal Longitudinal Study (NAPLS), includes three of the sites from the Lilly trial (Yale, UNC, and Toronto) and five additional sites (Harvard, Hillside, Emory, UCLA, and UCSD). Together, the group has a consolidated sample of over 400 prodromal patients, thus demonstrating that the "prodromal recruitment problem" is not insurmountable.

I just wanted to clarify a mistake in the article above. The new social and role functioning measures discussed by Barbara Cornblatt were incorrectly identified. The correct titles are Global Functioning Scale: Social (Auther et al., 2006) and Global Functioning Scale: Role (Niendam et al., 2006). Data on these new measures were recently published as part of a collaboration between LIJ and UCLA (Cornblatt et al., 2007). The references for these measures are listed below. Researchers interested in using these measures can contact either author (A. Auther or T. Niendam) for copies of the measures.

References:

Auther, A.M., Smith, C.W. & Cornblatt, B.A. (2006). Global Functioning: Social Scale (GF: Social). Glen Oaks, NY: Zucker Hillside Hospital.

Niendam, T.A., Bearden, C.E., Johnson, J.K. & Cannon, T.D. (2006). Global Functioning: Role Scale (GF: Role). Los Angeles: University of California, Los Angeles.

View all comments by Tara Niendam

The key issue of the confounding of the transition process and the related predictive analyses by uncontrolled treatment, especially with antipsychotic medications, has not been highlighted in the report. It would be illuminating to ask the collaborators to comment on this issue in the Forum. The randomized controlled trial conducted by Dr. McGlashan and colleagues (many of whom are now NAPLS investigators) was a stronger design since it contained a placebo arm which allows purer study of the prediction issue ( McGlashan et al., 2006). This should be supported in the future by NIMH and other funders in our opinion.

View all comments by Patrick McGorry
View all comments by Alison Yung

The case for testing antipsychotic drugs as prophylactic measures rests entirely on their empirically proven efficacy in decreasing the severity of positive psychotic symptoms among patients with established illness. Initial applications of these agents in studies of prodromal patients have produced discouraging results on the primary question of preventive effects. Among patients with established illness whose positive symptoms respond to antipsychotics, such treatment must be continuous in order to maintain treatment gains; it is therefore not surprising that trials of antipsychotics in prodromal patients would show effects of drug on positive symptom reduction only during the active treatment phase. With no demonstrable prophylactic effects, and with little or no effect on motivational symptoms or functional disability, antipsychotic drug treatment in the prodromal phase is clearly not the “silver bullet” of psychosis prevention.

Some have suggested that randomized controlled studies of antipsychotics provide the most coherent platform from which to monitor natural...  Read more

The case for testing antipsychotic drugs as prophylactic measures rests entirely on their empirically proven efficacy in decreasing the severity of positive psychotic symptoms among patients with established illness. Initial applications of these agents in studies of prodromal patients have produced discouraging results on the primary question of preventive effects. Among patients with established illness whose positive symptoms respond to antipsychotics, such treatment must be continuous in order to maintain treatment gains; it is therefore not surprising that trials of antipsychotics in prodromal patients would show effects of drug on positive symptom reduction only during the active treatment phase. With no demonstrable prophylactic effects, and with little or no effect on motivational symptoms or functional disability, antipsychotic drug treatment in the prodromal phase is clearly not the “silver bullet” of psychosis prevention.

Some have suggested that randomized controlled studies of antipsychotics provide the most coherent platform from which to monitor natural progression of the prodromal phase, since there is no confounding of progressive processes and treatment effects among individuals assigned to placebo. However, the ability to identify predictors and mechanisms of psychosis depends critically on the enrollment of large samples of prodromal subjects who are representative of the at-risk population. In the antipsychotic drug trials with prodromal patients, only a small fraction (on the order of 25 percent to 40 percent) of the potential subjects who screened as eligible for inclusion in the studies consented for participation and enrolled in the trials. Further, a relatively large fraction (on the order of 30 to 50 percent) of the initially enrolled subjects withdrew from these studies during the active treatment phase or follow-up period. Low rates of consent for participation combined with high attrition rates (likely due at least partially to side effects of drug treatments) have been reported in all prodromal randomized clinical trials and can thus be expected to severely constrain the generalizability of findings related to prediction and mechanisms of onset in such studies.

Therefore, ideally the next wave of studies into the natural progression of neurobiological factors in prodromal patients will employ a treatment algorithm by which antipsychotics are not prescribed until a patient displays positive symptoms at a fully psychotic level of intensity. This design will result in standardization of treatment across the participating sites and avoid confounding antipsychotic drug treatment with measurements of the biological processes hypothesized to underlie progression to psychosis.

Plump Enough
Thanks for your thought-provoking review of structural MRI changes in schizophrenia. I had a couple of quick comments.

You make the statement that, "Though cortical thickness itself is below the resolution of typical MRI, image analysis algorithms can now infer thickness across the entire cortical sheet as it winds its way throughout the brain." I thought sMRI gathers information for about 2 mm cubed or so. So maybe the point to make is that cortex thickness is not below the resolution, but the putative change in thickness is below the resolution. It would be interesting to know if the putative change in cortical thickness in schizophrenia could be better viewed with 3T or 7T scanners.

Also, I wonder how to interpret decreases in volume over five years that seem to be as much as 5 percent in some areas. How long could this continue to be progressive at this rate, and what would be the final cortical volume expected in the final decade of life? For example, if the DLPFC BA46 is about 3,500 microns thick, then a 5 percent loss/five years over 20...  Read more

Plump Enough
Thanks for your thought-provoking review of structural MRI changes in schizophrenia. I had a couple of quick comments.

You make the statement that, "Though cortical thickness itself is below the resolution of typical MRI, image analysis algorithms can now infer thickness across the entire cortical sheet as it winds its way throughout the brain." I thought sMRI gathers information for about 2 mm cubed or so. So maybe the point to make is that cortex thickness is not below the resolution, but the putative change in thickness is below the resolution. It would be interesting to know if the putative change in cortical thickness in schizophrenia could be better viewed with 3T or 7T scanners.

Also, I wonder how to interpret decreases in volume over five years that seem to be as much as 5 percent in some areas. How long could this continue to be progressive at this rate, and what would be the final cortical volume expected in the final decade of life? For example, if the DLPFC BA46 is about 3,500 microns thick, then a 5 percent loss/five years over 20 years would leave you with about 2,850 microns, and that would be about a 20 percent decrease in thickness. While postmortem studies may be limited, as Karoly points out, certainly we know that the frontal cortex is still "plump enough" to define cyto-architecturally, and to examine at the histological level. We also consider that there is about a 10 percent loss in cortical thickness in people with schizophrenia. Certainly, the cortex does not degenerate completely as would be expected with relentless progression of loss and accumulated deterioration of cortical grey matter over time.

Thus, this is an interesting issue, but many questions remain. Is there a lot of case-to-case variability that underlies these averages such that some cases lose more cortical volume and some do not lose any at all? Could it be that, while there is cortical volume loss, there are some patients in whom this loss slows or even reverses naturally over the course of the disease? What is the physical substrate of such cortical volume loss in people with schizophrenia? Can we prevent cortical volume loss over time, and would this be beneficial to patient outcomes?