Background

In mammals, the startle reflex is caused by a sudden and intense sensory stimulation. It is an evolved defensive reflex activity that can interrupt and interfere with ongoing cognitive and behavioral activities [1,2,3,4]. A gating mechanism can effectively inhibit the startle reflex, to ensure normal brain function. The prepulse inhibition (PPI) is the application of a weak prepulse stimulus that does not trigger a startle reflex during the first 30–500 ms before a strong stimulus. This interferes with and reduces the effects of the strong stimulus on the startle reflex [5]. PPI has good plasticity and has been widely used in various human and animal studies [6,7,8,9,10].

Previous studies have shown that PPI is regulated by the limbic-cortical-striatal-pallidal-thalamic (CSPT) neural circuit and the dopaminergic system [11, 12]. Further, it is found that injecting glutamatergic N-methyl-d-aspartate (NMDA) into the hippocampus can disrupt PPI by affecting the expression of Gamma-Aminobutyric Acid (GABA) neurotransmitters and the neural circuit [11, 13]. Braff et al. (1978) first found that patients diagnosed with schizophrenia had lower PPI than normal controls [14], and subsequent PPI research has focused on the schizophrenia spectrum population [15]. Moreover, PPI research with other mental disorders, including bipolar disorder (BD), obsessive-compulsive disorder, and autism spectrum disorders, has also been conducted [16,17,18], indicating that PPI abnormalities may be related to common psychopathological mechanisms of various disorders. PPI deficits have also been reported in first-degree relatives of patients with schizophrenia and BD, indicating that PPI may be a heritable phenotype [19, 20].

Although the neuromodulatory circuits of PPI are mainly at the brainstem level, many studies have confirmed that PPI is regulated by higher cognitive processes such as attention and emotion. Selective attention to prepulse stimulation can specifically enhance PPI in healthy subjects, while the enhancement effect in schizophrenia patients disappears due to attention deficit [21, 22]. Currently, some studies have shown that perceived spatial separation-induced PPI paradigm based on the priority effect can improve the individual recognition of prepulse sound and thus increase PPI [6]. Emotions play an important role in the selective attention and cognitive processes of people facing complex situations. For example, there can be higher PPI in response to pleasant or fearful pictures than to neutral pictures [23, 24]. Moreover, the CSPT circuit in BD patients may be impaired [25]. Deficits in this gating circuit could have adverse effects on cognitive information filtering, perhaps contributing to depressive thinking, manic thinking, and delusions.

Because of the complexity of BD, previous studies have drawn inconsistent conclusions. Some studies have found that, compared with healthy controls (HC), euthymic, or acute manic patients with BD, and first-degree relatives had significantly lower PPI levels [16, 20, 26, 27]. However, other studies found that patients with BD in the euthymic period and pediatric BD did not show PPI deficits [28,29,30]. There has been no systematic review or meta-analysis of PPI in patients with BD to assess the overall magnitude of these effects. The present study comprehensively retrieved the literature on PPI in patients with BD, then systematically reviewed the available data and conducted meta-analysis, comparing BD with HC.

Methods

Search strategy

Two authors independently searched relevant articles from the start of the database to July 27, 2019 in PubMed, Embase, the Cochrane Library databases, and the Chinese databases (VIPS, CNKI, and Wan Fang). Searches used medical subject headings (MESH), text words, and Boolean calculations. Main search terms included: bipolar disorder; bipolar affective disorder; mania; manic-depressive; manic episode; hypomania; hypomanic episode; bipolar depression; bipolar I disorder; bipolar II disorder; bipolar type I; bipolar type II; psychosis; psychoses; psychotic; prepulse inhibition; PPI; startle reflex; startle reaction; sensorimotor gating; and sensory gating. We also searched for relevant references in known quantitative papers, including non-English papers. The search strategy has been provided as supplementary material (Additional file 1). This review performed according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines, which provides an evidence-based minimum set of items for reporting in systematic reviews and meta-analyses. PRISMA Checklist has been included as supporting information “PRISMA checklist.doc” (Additional file 2).

Study selection criteria

Two authors independently screened the literature by reading the titles, abstracts, and the full text. The criteria for inclusion in the meta-analysis were: (1) Presence of a HC group. When participants included people with schizophrenia, BD, and first-degree relatives, information on patients with BD was extracted. (2) The paper contained PPI data or histograms and corresponding general demographic information with a PPI formula of PPI = 100 - [100*((prepulse - pulse amplitude) /pulse amplitude)]. (3) Presence of PPI data at 60 ms interstimulus interval (ISI). If the article contains both 60 ms and 120 ms PPI, 60 ms PPI data was extracted. (4) The quality of the research was evaluated by Newcastle-Ottawa Scale (NOS) [31]. This consists of nine items divided into three domains: selection of research subjects (four items); inter-group comparability (two items); and measurement of exposure factors (three items). The star system is used for semi-quantitative evaluation of research quality. The range of NOS is from zero to nine stars. Only papers with more than 4 stars were included in this study. Review papers, abstracts from conference proceedings and case reports were excluded. If the data were incomplete, the author was contacted by email to ask for information. If the author did not give a reply, we used the GetData software to extract the data in the paper. The two authors independently intercepted data to minimize errors in this study. Finally, we converted all standard error (SE) data into standard deviation (SD) form according to the formula (SD = SE× \( \sqrt{\mathrm{N}} \)). Due to incomplete data, studies that cannot be converted are included in the qualitative description. When the two authors reached different conclusions about eligibility, the third author was consulted.

Data analyses

Statistical analysis was performed using the software STATA 11.0 (Stata Corporation, USA). The heterogeneity test was evaluated using the I2 value. When I2 ≤ 50.0% and P ≥ 0.10, it indicated that the included studies are homogeneous, and the model of fix effect was adopted. If I2 is greater than 50.0% and P value is less than 0.10, results was considered heterogeneous, then the random effects models and sensitivity analyses were used [32]. Because the included studies used different continuous measures as outcome measures, we calculated effect sizes (Standard Mean Difference, SMD) and the standard errors of the effect sizes by the Cohen method in the metan command [33]. Galbraith graphs were plotted to test heterogeneity. Additionally, the metaninf command was used to examine the effect of individual studies on the total combined effect after removing outliers. Through the metabias command, the Egger’s test [34] was used to evaluate publication bias. The sample size, mean, SD, heterogeneity values (I2, p), SMD, and 95% confidence interval (CI) are shown in the forest map.

Results

Included studies

Figure 1 shows the article screening process: 815 unique articles were included in the initial screening, of which all but 16 were excluded after initial screening. Six more were excluded on further screening, leaving 10 papers included in qualitative synthesis, of which 5 were excluded from meta-analyses because of the absence of general demographic data, PPI data of 60 ms ISI and the maximum and minimum values required for data conversion. Therefore, five articles were included in final quantitative analysis for 60 mm ISI. The characteristics descriptions of all studies were displayed in Table 1.

Fig. 1
figure 1

Literature search flow

Full size image
Table 1 Studies included in systematic review
Full size table

Meta analyses

A total of 141 euthymic patients and 132 HC were included in the meta-analysis of the 60 ms ISI. BD diagnoses in all 5 studies were according to Structured Clinical Interview for DSM Disorders-Fourth Edition (DSM-IV) criteria. The clinical characteristics and treatment of patients with BD are depicted in Table 2.

Table 2 Studies included in the meta-analysis
Full size table

Furthermore, as shown in Table 2, in all five studies the PPI was calculated the same way and the mood state of patients with BD is euthymic. Therefore, we performed meta-analysis for PPI of 60 ms ISI. As can be seen in Fig. 2, there was an overall PPI difference between euthymic BD and HC at 60 ms ISI (SMD = − 0.32, 95% CI = − 0.54 – -0.10).

Fig. 2
figure 2

Forest plot of effect size with random effect model between BD patients and healthy controls for PPI in 60 ms interstimulus interval. The vertical solid line (0 on the abscissa scale) represents the invalid line. The dotted line represents the point estimate of all the data summarized

Full size image

Quality assessments

Using the NOS to assess the quality of the studies, all 5 papers gained 4 stars or more (Table 3).

Table 3 The quality assessment by NOS scale
Full size table

Heterogeneity and sensitivity analyses

As shown in Fig. 2, the PPI of the 60 ms ISI had little heterogeneity (P = 0.476, I2 = 0.0%, SMD = − 0.32, 95% CI = − 0.54 - -0.10). According to the heterogeneity analysis results for 60 ms PPI, all research points fall between two regression lines (Fig. 3). Sensitivity analysis shows no significant change in the combined effect value after removing this single study (Fig. 4). Due to lack of data on BD I or II subtypes, subgroup analyses of BD type cannot be obtained.

Fig. 3
figure 3

The Galbraith plot for 60 ms PPI heterogeneity. There are three diagonal lines in the figure, the middle diagonal line represents combined value of the fixed effect, and the 95% confidence interval on both sides

Full size image
Fig. 4
figure 4

Sensitivity analysis for 60 ms PPI. The middle vertical line in the figure is a vertical line with a difference value of − 0.32, which is the total combined effect size. The points corresponding to each study represent the combined effect size of the remaining studies after deletion of the study

Full size image

Publication bias

When using the Egger’s test, there was no publication bias at 60 ms (P = 0.606).

Discussion

To the best of our knowledge, this is the first systematic review and meta-analysis about PPI levels in patients with BD. The main finding of this study is that euthymic patients with BD had significant PPI deficits compared with HC when the ISI between pulse and prepulse was 60 ms. In the PPI paradigm of human studies, the ISI generally range from between 30 and 240 ms. It was found that maximum PPI occurs at 60-120 ms ISI in schizophrenia patients and normal subjects [14, 39, 40]. Even if BD are treated with an mood stabilizers, atypical antipsychotics, or have a history of smoking, all of which can affect PPI, patients still have PPI deficits at 60 ms ISI. This suggests that BD is more associated with a deficit in the sensory gate at 60 ms ISI which is an insensitive interval for drugs that affect sensory gating [41,42,43].

The PPI at 120 ms ISI is regulated by attention and distribution in advanced cognitive function and is susceptible to drug effects [21]. A total of 10 studies explored PPI levels at 120 ms ISI in patients with BD and the conclusions are different. Four studies reported PPI defects of patients with BD and 4 studies showed no defects, other two studies explored gender differences in PPI and did not show differences between overall patients and HC. Currently, the studies include those with BD in euthymic, and mania/mixed states, as well as those with BD I or BD II, and it includes those with adults or children. The clinical heterogeneity of the 120 ms ISI in BD patients was more evident, so no meta-analysis was applied for 120 ISI. However, at present, clinical heterogeneity is difficult to explain by analyzing data. Just for the heterogeneity depend on the current analyses, further study is needed with good design.

There are multiple definitions of the euthymic period in the literature reviewed. One study suggested that euthymia was defined as score below 7 on the Hamilton Depression Rating Score (HDRS) and below 6 on the Young Mania Rating Scale (YMRS) [16]. Another used criteria of less than 8 on Hamilton Depression Rating Scale (HAMD) and less than 20 on YMRS [28]. Another included euthymic (HAMD < 7 and YMRS < 7) patients with BD [20]. Yet another study relied on self-report [26]. However, despite inconsistent criteria, by these criteria all participants were considered clinically stable at time of testing. Euthymic patients with BD had PPI deficits at 60 ms, suggesting a consistent association between BD and sensory gating deficit. This is consistent with previous studies, euthymic patients with BD still have functional deficits of sustained attention and find it difficult to ignore the irrelevant stimulus [44, 45]. At present, there is little research on PPI in the acute phase of BD and further research is needed. PPI deficits may be due to both trait and state characteristics.

Schizophrenia-related studies have shown that PPI deficits are strongly associated with positive symptoms [9, 46]. A meta-analysis of the previous P50 showed that patients with BD have sensory gate deficits, which become more serious in when exhibiting psychotic symptoms [47]. One study found that psychotic patients with BD in the manic phase have significant PPI deficits [27]. Another study of manic patients without psychotic symptoms did not find this [29]. However, Sanchez-Morla et al. support the presence of PPI deficits in stable patients, which have no association with psychotic symptoms [16]. Since PPI is found in other non-psychotic disorders and is regulated by attention, patients with BD have sustained attention deficit in different stages of the disease, so it may be a trait characteristic of BD. Unfortunately, due to insufficient PPI data, we were unable to do the relevant subgroup analysis. The relationship with the psychotic symptoms of patients with BD remains to be further studied.

PPI deficits were first demonstrated in patients with schizophrenia and have been studied primarily in patients with schizophrenia spectrum disorders [14, 15]. This meta-analysis adds to the increasing evidence that patients with BD also have impaired PPI levels, suggesting that PPI deficits are not unique to schizophrenia. However, there is overlap of symptoms and genetic communalities between BD and schizophrenia [48]. It is possible that the sensory attentional deficit assessed by PPI involves some specific neurological mechanism implicated in both disorders. BD and schizophrenia in current diagnostic systems may be different manifestations of very similar underlying brain dysfunctions and clinical phenotypes of psychiatric disorder. This is therefore a promising direction for researching ways of understanding and classifying psychiatric disorders by their biological causes, rather than by their symptom clusters.

There are some limitations of this study. PPI is affected by several types of factors. Firstly, among the literatures included in the quantitative analysis, patients in three literatures were not described the type of BD (bipolar I or II), and all the patients in the other two researches were BD I. An effective subgroup analysis cannot be performed. Secondly, according to the HAMD score, Gogos 2009 included those who were depressed assessed by HAMD [26]. The group mean is in the mild depression range, with the SD indicating that there were quite a few participants who were likely mildly depressed, but self-reported euthymic. Thirdly, PPI is affected by many factors. A number of previous studies found that the second generation of antipsychotics can improve PPI, and that mood stabilizers such as, lithium and valproate, can increase PPI [49,50,51]. Previously in the study of BD, most patients were prescribed drugs and had multiple drug combinations, which cannot be easily converted to equivalent dosages. PPI levels are also affected by gender, age, smoking, and other factors. Most studies suggest that male PPI levels are significantly higher than female [52, 53]. Kumari et al. suggested that the level of PPI in minors is significantly lower than in adult patients [54]. nicotine can also improve PPI levels [55]. These factors are hard to control in research and may be causes of heterogeneity. Furthermore, to carry out quantitative analysis, we use data interception software to capture data from pictures in previous literature, resulting in no deviation results between the data used and the original data.

Conclusions

In conclusion, the current systematic review and meta-analysis suggests that BD patients show PPI deficits in the 60 ms ISI. Further research on PPI in disorders other than schizophrenia is needed, including BD during the acute phase and psychotic state, using consistent criteria for defining euthymic, acute phases, and psychotic state. More research is needed in the future to confirm this outcome, and to delve deeper into the mechanisms behind deficits.