Meta-analysis of mismatch negativity to simple versus complex deviants in schizophrenia

Abstract

Mismatch negativity (MMN) deficits in schizophrenia (SCZ) have been studied extensively since the early 1990s, with the vast majority of studies using simple auditory oddball task deviants that vary in a single acoustic dimension such as pitch or duration. There has been a growing interest in using more complex deviants that violate more abstract rules to probe higher order cognitive deficits. It is still unclear how sensory processing deficits compare to and contribute to higher order cognitive dysfunction, which can be investigated with later attention-dependent auditory event-related potential (ERP) components such as a subcomponent of P300, P3b. In this meta-analysis, we compared MMN deficits in SCZ using simple deviants to more complex deviants. We also pooled studies that measured MMN and P3b in the same study sample and examined the relationship between MMN and P3b deficits within study samples. Our analysis reveals that, to date, studies using simple deviants demonstrate larger deficits than those using complex deviants, with effect sizes in the range of moderate to large. The difference in effect sizes between deviant types was reduced significantly when accounting for magnitude of MMN measured in healthy controls. P3b deficits, while large, were only modestly greater than MMN deficits (d=0.21). Taken together, our findings suggest that MMN to simple deviants may still be optimal as a biomarker for SCZ and that sensory processing dysfunction contributes significantly to MMN deficit and disease pathophysiology.

Keywords: Complex deviant; Meta-analysis; Mismatch negativity; P300; P3b; Schizophrenia.

Figure 1

MMN deficits in SCZ by deviant category. The overall pooled effect size calculated using a fixed effects model is shown with 95% confidence intervals (horizontal error bars) for each simple deviant category. Exact values are reported on the right column including confidence intervals in brackets. Longer and shorter duration deviants were also analyzed separately. Positive values indicate lower MMN magnitudes in SCZ.

Figure 2

MMN deficits in SCZ using complex sensory deviants. Effect sizes for individual HC vs SCZ study samples are shown for studies using complex sensory deviants (as defined in Section 2.2) with 95% confidence intervals plotted as horizontal error bars. The pooled effect size using a fixed effects model is shown at the bottom including its 95% confidence interval. Positive values indicate lower MMN magnitudes in SCZ.

Figure 3

MMN deficits in SCZ using complex abstract/pattern deviants. Effect sizes for individual HC vs SCZ samples are shown for studies using complex abstract/pattern deviants (as defined in Section 2.2) with 95% confidence intervals plotted as horizontal error bars. The pooled effect size using a fixed effects model is shown at the bottom including its 95% confidence interval. Positive values indicate lower MMN magnitudes in SCZ.

Figure 4

A comparison of MMN and P3b deficits in SCZ. An effect size representing the difference in effect sizes for MMN and P3b in SCZ vs HC is shown for individual studies that measured both ERP components. 95% confidence are depicted as horizontal error bars. Exact values are reported on the right column including 95% confidence intervals in brackets. Positive values indicate greater deficits for SCZ in P3b than in MMN.

Figure 5

P3b deficits correlate with MMN deficits in SCZ. The effect size of P3b deficits is plotted against the effect size of MMN deficits in each SCZ vs HC sample in which both components were measured. The dashed line represents the results of a correlation between the two variables computed from a weighted linear regression. Symbol sizes depict the weight given to each sample as determined by the variance estimate of MMN and P3b effect sizes. R is the correlation coefficient after adjusting for age and sex.

Figure 6

The relationship between MMN deficits in SCZ and MMN magnitude in HC. (A) The effect size for SCZ vs HC sample comparison is plotted against the normalized mean of MMN in HC. Symbol colors denote the different deviant type categories (see key). The solid line is an unweighted linear regression. (B) Unweighted means for the effect sizes shown in A by deviant type as filled circles (duration also shown as longer and shorter duration separately). Empty circles represent the means adjusted for normalized amplitude of MMN in HC and error bars represent SD.