Mismatch Negativity in Han Chinese Patients with Schizophrenia: A Meta-Analysis.

BACKGROUND: Previous meta-analysis revealed that mismatch negativity(MMN) amplitude decreased in patients with schizophrenia compared with healthy controls (Cohen's d, d about 1), leading to the possibility of mismatch negativity being used as a biomarker for schizophrenia. However, it is unknown whether MMN is reliably changed in Chinese patients. It is necessary to carry out a meta-analysis on MMN of Han Chinese patients with schizophrenia. AIM: To investigate whether MMN could be used as a biomarker for Han Chinese patients with schizophrenia.
METHODS: A literature search was conducted to identify clinical trials on MMN in Han Chinese schizophrenia patients published before May 8, 2017, by searching the Chinese language databases CNKI, WanFang Data, VIP Data and PubMed. The effects of MMN deficits were evaluated for MMN amplitude by calculating standard mean difference (SMDs) between schizophrenia patient groups and healthy control groups.
RESULTS: A total of 11 studies were included in the analysis. The total quality of all the studies were more than 6 as evaluated by Newcastle-Ottawa Scale (NOS). Meta-analysis of data from these studies had a pooled sample of 432 patients with schizophrenia and 392 healthy controls. There exists significant MMN deficit in schizophrenia patients compared to healthy controls (Cohen's d=1.004). When studies were excluded due to heterogeneity, the pooled effect size of the MMN differences between the patient group and healthy controls dropped to 0.79 (Cohen's d=0.79). Subgroup analysis showed that MMN amplitude deficits of schizophrenia over three years had the pooled effect size of 0.95, and less than three years had the pooled effect size of 0.77. Publication bias conducted via Egger regression test (t = 1.83; p = 0.101), suggested that there was no publication bias.
CONCLUSION: The effect size of MMN amplitude between Chinese patients with schizophrenia and healthy controls is consistent with other meta-analyses published on this topic, suggesting that Han Chinese patients with schizophrenia also exhibited MMN deficits.

1. Introduction

It is well documented that cognitive impairment is a common feature of schizophrenia. A variety of cognitive abnormalities in brain function and structure are detected in patients with schizophrenia.[ Hence, the implementation of validated biomarkers would improve the accurate diagnosis of cognitive dysfunction in schizophrenia patients. Mismatch negativity (MMN) is obtained by subtracting the event-related potential (ERP) to a deviant stimulus in a sequence of standard stimuli, elicited by auditory oddball paradigms. MMN is the negative component, typically peaking at about 250 ms from change to onset, which reflects the ability of actively and accurately predicting forthcoming sensations in the brain.[ MMN is most commonly evoked using auditory stimuli that reflect information processing and encoding in the auditory cortex. Unlike the P300 test, in which the subject is actively reacting to the presence of target stimuli, the processes involved in MMN generation are assumed to initiate involuntary auditory change detection.[ Thus, the mismatch negativity component could be considered as a hallmark of auditory prediction errors [ and is well described as the electrophysiological response.[

MMN amplitude has been consistently shown to be attenuated in schizophrenia compared to healthy controls. In large-scale multi-site clinical studies from the US,[ highly significant MMN (Cohen’s d, d=0.96) amplitude reductions were observed in SZ patients, it is so far probably the most promising neurophysiological biomarkers to be used to guide assignment of patients to clinical interventions. Systematic reviews and meta-analysis reveal that reliability coefficients and effect sizes of deficits in schizophrenia for MMN are stable in distinguishing healthy controls from those with schizophrenia.[ Furthermore, impaired MMN is closely associated with N-methyl D-aspartate receptor hypo-functioning which exist as race-related biological differences.[ However, it is remains unclear whether Han Chinese individuals with schizophrenia show altered MMN responses and therefore we performed a meta-analysis of peer-reviewed MMN studies that had specifically targeted Han Chinese schizophrenia patients to examine the pooled effect size. This may contribute to the view of MMN as a translatable biomarker for diagnosing schizophrenia in China.

2. Methods

2.1 Criteria for considering studies

Study types considered for meta-analysis included all relevant parallel control studies on MMN with individuals who had a diagnosis of schizophrenia and healthy controls. Types of outcome measures included: (a) The study should include at least one psychiatrically healthy control group and individuals with schizophrenia. (b) Studies that have been carried out with a Han Chinese sample. (c) The MMN amplitude must be reported as a difference wave. (d) MMN amplitude must be extracted from auditory oddball paradigms, which have been clearly introduced. Types of outcome measures excluded: (a) Conference paper, academic dissertation, reviews or papers that didn’t provide data; (b) Redundant publications; (c) other paradigms, e.g. paradigm (visual), or those which have not been clearly introduced.

2.2 Search methods for identification of studies

A literature search for papers published before May 8, 2017 was conducted using the following electronic databases: CNKI, Wanfang DATA, VIP and Pubmed. Keywords included synonyms of “schizophrenia” or “psychosis”, “Chinese” or “China”, “mismatch negativity” or “MMN” in PubMed (and the Chinese equivalents in Chinese databases). A set of search terms were ((“schizophrenia”[MeSH Terms] OR “schizophrenia”[All Fields]) OR (“psychotic disorders”[MeSH Terms] OR (“psychotic”[All Fields] AND “disorders”[All Fields]) OR “psychotic disorders”[All Fields] OR “psychosis”[All Fields])) AND ((mismatch[All Fields] AND negativity[All Fields]) OR MMN[All Fields]) AND ((“asian continental ancestry group”[MeSH Terms] OR (“asian”[All Fields] AND “continental”[All Fields] AND “ancestry”[All Fields] AND “group”[All Fields]) OR “asian continental ancestry group”[All Fields] OR “chinese”[All Fields]) OR (“china”[MeSH Terms] OR “china”[All Fields])) were utilized.

2.3 Outcome measures

The outcome measures were MMN amplitude between the schizophrenia group and healthy control group. The effect size between the two groups was calculated using the difference value of MMN amplitude. The larger the effect size, the more serious the MMN amplitude deficits in schizophrenia, and the greater the difference between the two groups.

2.4 Data collection and analysis

2.4.1 Selection of trials

Two researchers(XYB and ZL) independently conducted a full-text review, extracted data cross checks, and consulted a third author (LXB) if there were any inconsistencies.

2.4.2 Data collection

Both authors independently extracted the data from the included trials. Any disagreement was discussed, the decisions documented. The remaining problems were arbitrated by the third reviewer (LXB).

2.4.3 Data synthesis

Meta-analysis: Unbiased Cohen’s effect size (d) for each study was calculated by MMN amplitude as previously described. Stata 11.0 software was used to calculate size of effect from each study, and the average (d+) means d < 0.2 is a negligible effect size, d > 0.2 is a small effect size, d > 0.5 is a medium effect size, d > 0.8 is a large effect size.[

Heterogeneity: Study heterogeneity was quantified for the outcome analysis using the I2 statistic alongside the Chi2 ‘P’ value with I2 ≥ 50% indicating significant heterogeneity. When heterogeneity is present, sensitivity analyses was conducted to assess potential influences of any one single study on the pooled MD and associated P-values,[ and we conducted a subgroup analyses by the course of schizophrenia, to explore whether MMN deficits were associated with the course of illness in schizophrenia.

The quality assessment of meta-analysis outcomes: Newcastle-Ottawa Scale was used for assessing the methodological quality of each study. The tool is used for assessing the quality of case-control and cohort studies. It is judged on eight items, categorized into three groups: the selection of the study groups; the comparability of the groups; and the ascertainment of either the exposure or outcome of interest for case-control or cohort studies respectively. The total score was 10. The higher quality studies are given higher scores.[

Assessment of publication biases: To evaluate reporting bias, effect size data were inspected visually using funnel plots. The theory underlying this approach is that small studies have greater standard errors.[ This was investigated using a funnel plot that is symmetrical if there is no bias. The degree of funnel-plot asymmetry was evaluated using the Egger regression test.[ To provide an estimation of the possible effect of unpublished studies, effect sizes were corrected using the “trim and fill” method.[

3. Results

3.1 Results of the search

This initial search strategy identified 63 articles on MMN in Han Chinese patients with schizophrenia, 56 of which were published in Chinese and 7 of which were in English. 18 studies met the criteria mentioned above after duplicate checking via EndNote 7X. 11 unique articles were selected for inclusion in the meta-analysis. 8 of these studies were in Chinese and 3 were in English (Fig. 1). 11 studies were included and the final total study population (total 824) consisted of 432 schizophrenia patients and 392 healthy controls. The combined demographic information from these studies is presented in Table 1.

Figure 1.

Flowchart of literature search and exclusion process

Table 1.

Characteristics of included studies

Study	Stimulus type	Standard/Hz	Deviant/Hz	Age/year SZ/HC	Course	Sample size		Amplitude (SZ)		Amplitude (HC)		Effect size
						SZ	HC	Mean	SD	Mean	SD
Lv2007	frequency	500	2000	31.4/33.4	<1.6Y	43	36	-4.6	3.3	-7.9	3.6	0.96
Lin2012	frequency	500	1000	36.2/38.0	N*	62	46	-0.4	0.8	-1.0	1.0	0.69
Li2015	frequency	1000	2000	26.8/25.8	10.51M	45	50	-5.8	1.2	-6.5	1.1	0.54
Sun2014	frequency	500	2000	32/33	5.4 Y	78	62	-5.7	3.2	-8.4	3.5	0.81
Yuan2009	frequency	1000	1500	34/34	N*	23	23	-1.6	0.9	-2.0	0.4	0.63
Yang2012	frequency	1000	2000	30.1/28.2	N*	31	30	-1.9	0.8	-3.4	1.2	1.50
Yao2012	frequency	1000	1500	37/N*	19 M	30	30	-1.5	0.9	-2.1	0.2	0.92
Zhang2012	frequency	1000	2000	28.6/29.1	N*	35	30	-3.7	0.9	-4.9	1.6	0.92
Zhao2012	frequency	1000	2000	28.6/30.4	>5Y	21	26	-4.0	2.1	-4.5	2.0	0.27
Zhou2013	frequency	1000	1500	33/33	N*	26	26	-0.6	0.9	-2.1	0.5	2.52
Zha2015	frequency	500	2000	29.7/29.3	38 W	36	33	-3.3	1.1	-6.8	2.3	1.97

*N means not mentioned, SZ=schizophrenia group, HC=healthy controls group. Y=years, M=months, W=weeks

3.2 Quality assessment

According to the Newcastle-Ottawa Scale, all the included studies have clear diagnostic criteria for schizophrenia, and the healthy controls group and schizophrenia group had no statistical difference in age and gender, and all the healthy controls had no family history of mental illness. Therefore, the two groups are comparable. In comparable education, there were 4 studies that didn’t match between groups (Lin 2012 [, Yuan 2009 [, Zhao 2012 [, Zhou 2013 [). All the included studies didn’t mention the non-response rate for both groups. The total quality of all the studies was more than 6, suggesting that the quality of included studies was high.[ See Table 2.

Table 2.

The methodological assessment of case control trials by Newcastle-Ottawa Scale

Study	Selection			Comparability			Exposure			Total Quality score
	1*	2*	3*	4*	5*	6*	7*	8*	9*
Lv 2007[28]	1	1	1	1	1	1	1	1	0	8
Lin 2012[19]	1	1	1	1	1	1	1	1	0	7
Li 2105[29]	1	1	1	1	1	1	1	1	0	8
Sun 2014[30]	1	1	1	1	1	1	1	1	0	8
Yuan 2009[20]	1	1	1	1	0	1	1	1	0	7
Yang 2012[25]	1	1	1	1	1	1	1	1	0	8
Yao 2012[31]	1	1	1	1	1	1	1	1	0	8
Zhang 2012[24]	1	1	1	1	1	1	1	1	0	8
Zhao 2012[21]	1	1	1	1	0	1	1	1	0	7
Zhou 2013[22]	1	1	1	1	1	1	1	1	0	7
Zha 2015[23]	1	1	1	1	1	1	1	1	0	8

*1=The case definition is adequate with independent validation; 2=Representativeness of cases; 3=Community Controls; 4=Controls with no history of disease; 5=Cases and controls with comparable education; 6=Cases and controls with comparable age; 7=Ascertainment of outcome; 8=Same method of ascertainment for cases and controls; 9=Similar non-response rate for both groups.

3.3 Meta analysis

3.3.1 MMN Amplitude effect size

Heterogeneity was identified across all studies determined by Q-statistics and Galbraith index (I2=75.3%, df=10, p<0.001), indicating significant heterogeneity. Fig. 2 shows the effect size and 95% confidence interval for each investigation. The pooled effect size (random model) of the MMN differences between schizophrenia patients and healthy controls was 1.004 with 95%CI ranging from 0.703 to 1.305.

Figure 2.

Effects of MMN amplitude in Patients with Schizophrenia (SZ) vs Healthy Controls (HC); The x axis represents effect size

3.3.2 Sensitivity analysis

Q-test revealed that heterogeneity has been identified across all studies (Stata11.0, I2=75.3%, df=10, p<0.001). Sensitivity analysis showed 2 studies (Zha 2015[, Zhou 2013[, et al) had significant heterogeneity from the others. When we excluded these 2 studies, Q-test was not statistically significant (I2=33.1%, df=7, p=0.153), However, the d+ value was reduced to 0.79 (0.60, 0.99), Fig. 3. It is indicated that these studies are the reason for heterogeneity.

Figure 3.

Effects of MMN amplitude in Patients with Schizophrenia (SZ) vs Healthy Controls (HC) when excluding two studies; The x axis represents effect size

3.3.3 Subgroup analysis

In order to assess if there is a relationship between the course of disease and MMN impairment across studies, we performed subgroup analysis. The schizophrenia patients were separated into groups with two disease stages: one group was over 3 years and another was below 3 years. Effect sizes by group are depicted in Fig. 4. 5 studies were excluded due unknown course of illness records (Zhang 2012[, Yang 2012[, Lin 2012[, Yuan 2009[, Zhou 2013[). 6 studies were included in the subgroup analysis. Among individuals who had been diagnosed with schizophrenia for more than three years, the results were heterogeneous across groups (I2=89.3%, df=2, p<0.001) and the pooled effect size was 0.95(0.68, 1.21), and less than three years had the pooled effect size of 0.77. For the patients who had schizophrenia for less than three years, the results were homogenous (I2=5.9%, df=2, p=0.345) and the pooled effect size was 0.77 (95%CI: 0.51, 1.04).

Figure 4.

Effects of MMN amplitude in Patients with Schizophrenia (SZ) vs Healthy Controls (HC) by subgroup analysis with the course of schizophrenia when over 3 years; The x axis represents effect size

3.3.4 Publication bias

Plotting MMN amplitude effect size from each study in a funnel plot revealed symmetry, suggesting that the sample could not be affected by reporting bias (Egger regression test: t = 1.83; p = 0.101; 95% CI, -1.17 to 11.08).

4. Discussion

4.1 Main findings

From this paper, 11 studies were included for performing meta-analysis of MMN in Han Chinese patients with schizophrenia. MMN amplitude also decreased in Han Chinese patients with schizophrenia, and the pooled effect size was large (1.004), close to 0.95-0.96, which are the effects sizes seen in other published metaanalyses.[ This result demonstrates that the effect size of amplitude in Han Chinese schizophrenia is similar to non-Han Chinese schizophrenia. It also suggested MMN could be a reliable biomarker for monitoring cognitive function alteration in Han Chinese patients with schizophrenia, as it is not affected by ethnic factors. However, the heterogeneity test revealed that there exists heterogeneity across all studies. 2 studies (Zha 2015[, Zhou 2013[, et al) showed significant heterogeneity from the others, after excluding these studies, the average value of effect size (d+) decreased to 0.79, demonstrating that these 2 studies raised the overall effect sizes.

Previous study showed that first-episode schizophrenia patients had an average effect size of 0.42, far less than chronic schizophrenia patients. Moreover, MMN attenuation in first-episode schizophrenia patients was significantly lower than in chronic schizophrenia patients. The MMN frequency appeared to be oppositely attenuated along with later disease stage. Thus, MMN amplitude might reflect disease progression.

In order to investigate whether MMN deficit was correlated with the progression of schizophrenia, we performed subgroup analysis by dividing recruited schizophrenia patients into two groups, depending on whether the onset of schizophrenia was more than three years. Usually, schizophrenia patients with three year disease duration or above are considered chronic patients.[ Patients with schizophrenia over three years had the pooled effect size of 0.95, and patients with schizophrenia less than three years had the pooled effect size of 0.77. Therefore, MMN amplitudes were slightly lower in patients with schizophrenia below three years compared to those with schizophrenia over three years, suggesting that MMN amplitude may reflect disease progression. These findings suggest frequency MMN might be a state marker of schizophrenia.

4.2 Limitations

Some methodological limitations of the current study should be taken into account. First, in terms of Chinese database retrieval, we did not search the databases from Hong Kong, Macao and Taiwan. Second, in terms of English database retrieval, we did not search database such as EMBASE and PsycInfo. This may have caused incomplete retrieval of studies. Third, we did not investigate the correlation between MMN amplitude and treatment, therefore, we do not know whether MMN can be used as an indicator of efficacy evaluation for antipsychotic drugs. Finally, the types of MMN in the included studies are all frequency MMN, which cannot be compared between different kinds of MMN effect values.

4.3 Implications

In summary, MMN amplitude deficits in Han Chinese schizophrenia are similar to the reports from western countries. Consistently, there also exists a large effect size of MMN between Chinese schizophrenia patients and healthy controls, which may be used as a biomarker of schizophrenia. At present, most of the studies on schizophrenia in China are caused by a single frequency deviation stimulus. In the future, a control study of MMN caused by different stimuli can be carried out. The current research focused on analyzing amplitude of MMN, in-depth analysis of MMN data, such as duration-frequency analysis, may be a new research approach in the future. It is also necessary to carry out multi-center MMN experiments to establish MMN norms for patients with schizophrenia in China.