Fifty adults participated voluntarily in the present study and were selected from a larger sample referred to our research group from Primary Care Health Centres in Santiago de Compostela, Galicia (Spain). The participants were divided into three groups: (1) Control, comprising 20 adults (mean age: 65.8 years) with normal cognitive functioning; (2) sdaMCI, comprising 19 adults (mean age: 67 years); and (3) mdaMCI, comprising 11 adults (mean age: 71 years). The three groups were matched regarding age and years of education. The demographic and neuropsychological measures of the three groups, and the differences between groups (as determined by the corresponding analysis) are shown in Table 1.

The present study complied with the ethical standards established in the 1964 Declaration of Helsinki (Lynöe et al., 1991) and was approved by the Galician Clinical Research Ethics Committee (Xunta de Galicia, Spain). All participants gave their written informed consent to participate in the study, after the procedures were fully explained to them. Participants did not report any medical or psychiatric diseases, history of clinical stroke, traumatic brain injury, motor-sensory deficits or substance abuse/dependence (alcohol or drug). Adults with scores higher than 10 in depression screening (Geriatric Depression Scale, GDS; Yesavage et al., 1983) were excluded. All participants were right-handed, as evaluated by the Edinburgh inventory (Oldfield, 1971), and all had normal audition and normal or corrected-to-normal vision.

All participants then underwent the following neuropsychological tests: (1) the Spanish version of the Mini-Mental State Examination (MMSE; Folstein et al., 1975; Spanish version by Lobo et al., 1999); (2) the Spanish version of the Californian Verbal Learning Test (CVLT; Delis et al., 1987; Spanish version, TAVEC by Benedet and Alexandre, 1998), which evaluates short-delay and long-delay free recall, as well as short-delay and long-delay recall with semantic cues; and (3) the Spanish version of the Cambridge Cognitive Examination (CAMCOG-R), which assesses impairments in perception, attention-calculation, praxis, language and executive functioning (Huppert et al., 1996).

When all participants were correctly diagnosed and classified as Control, sdaMCI or mdaMCI, they voluntarily attended our laboratory for psychophysiological evaluation by the ERP technique. Participants categorized as having sdaMCI or mdaMCI met the criteria for MCI outlined by Albert et al. (2011) and the criteria for aMCI proposed by Petersen (2004). All aMCI adults were characterized by self-reported memory complaints corroborated by an informant, performance of less than 1.5 SDs below age norms for the CVLT, essentially preserved activities of daily living, and no dementia. In addition, participants with mdaMCI scored less than 1.5 SDs below age- and education-related norms in the MMSE and at least two cognitive subscales of the Spanish version of the CAMCOG-R.

All control adults scored higher than the cut-off on memory, general cognitive functioning, and specific cognitive domain tests. For a more detailed description of the global samples, the criteria for inclusion/exclusion, the tests used, and the diagnosis and classification criteria, see Juncos-Rabadán et al. (2013).

Participants performed an auditory-visual distraction-attention task, which was an adaptation of the task used by Escera et al. (1998, 2001). Five hundred auditory-visual (A-V) pairs of stimuli (divided in two blocks) were presented. In each A-V pair, the duration of the auditory stimulus was 150 ms and the duration of the visual stimulus was 200 ms (onset-to-onset interval: 300 ms). An onset-to-onset interval between pairs of stimuli of 2-s duration was used. Participants were instructed to pay attention to the visual stimuli and to ignore the auditory stimuli. See diagram of the task procedure in Figure 1.

The unattended auditory stimuli, of intensity 75 dB SPL, were presented binaurally, via headphones. Three types of auditory stimuli were presented: a standard tone of 1000 Hz (presentation probability of 70%), a deviant tone of 2000 Hz (presentation probability of 15%), and novel stimuli (different each time, e.g., glass crashing, phone ringing; presentation probability of 15%). Three types of attended visual stimuli were also presented: numbers (2, 4, 6, and 8), with a presentation probability of 33%, letters (a, c, e, and u), with a presentation probability of 33%, and triangles (pointing right, left, upward or downward), with a presentation probability of 34%. Participants were instructed to respond (Go condition) to the numbers and to the letters by pressing two different buttons, each with a different hand (the buttons were counterbalanced among participants) and to inhibit their responses to triangles (NoGo condition).

Recording sessions were conducted in a Faraday chamber, under attenuated levels of light and noise. The participants were seated on a comfortable chair during the electroencephalographic (EEG) recording and they were instructed to move as little as possible during the session.

Visual stimuli were presented with a subtended visual angle of 1.7 × 3.3° of arc, on a 19″ flat screen monitor (located at a distance of one meter from the participant), with a vertical refresh rate of 120 Hz. Forty-nine ring electrodes (placed in an Easycap GmbH, according to the International 10–10 system) were used for the EEG recording. An electrode placed at the tip of the nose served as the reference for all active electrodes, and an electrode at Fpz served as ground. Two electrodes placed at the outer canthi of both eyes were used to record the horizontal electro-oculogram (HEOG), whereas two electrodes placed near the right eye (one supra-orbitally and other infra-orbitally) were used to record the vertical electro-oculogram (VEOG). The EEG signal was passed through a 0.01–100 Hz analog bandpass filter and sampled at 500 Hz. The impedances were below 10 kΩ.

Off-line, ocular artifacts were corrected (using the method Gratton et al., 1983) and the EEG epochs of -150–1300 ms synchronized with each auditory stimulus (standard, deviant, or novel) were extracted (epochs in Figures 3, 4, 5A were shortened from -150 to 800 ms). Finally, three conditions were evaluated: Standard (standard auditory stimulus-Go visual stimulus), Deviant (deviant auditory stimulus-Go visual stimulus), and Novel (novel auditory stimulus-Go visual stimulus). At least 50 artifact-free epochs were obtained for each condition. Epochs related to the auditory stimuli-NoGo visual stimuli were not evaluated.

The ERP waveforms obtained for the three groups (Control, sdaMCI, and mdaMCI) in each condition are shown in Figure 3, and the ERP waveforms obtained for the three conditions (Standard, Deviant, and Novel) in each group are shown in Figure 4.

Prior to averaging, the signal was filtered using a 0.1–30 Hz (24 dB/octave slope) digital bandpass filter. In addition, the epochs were corrected to the mean voltage of the 150-ms pre-stimulus recording period, and those with signals exceeding ±100 μV, as well as the first five epochs of each block, were rejected.

Finally, N-S difference traces were obtained (see Figure 5A) with the aim of identifying and measuring P3a.

The percentage of correct responses (hits) and RTs (between the onset of the Go visual stimulus and pressing the key) were evaluated for each group (Figure 2).

Two phases were identified in the temporal range of P3a, for the control, sdaMCI and mdaMCI participants (see Figures 3, 4, 5A): (1) early P3a (e-P3a), in a latency range between 280 and 400 ms post-stimulus and with a maximum amplitude at fronto-central locations, and (2) late P3a (l-P3a), in a latency range between 350 and 500 ms post-stimulus, and with a maximum amplitude at parieto-central locations. A scatter plot of individual dates for the amplitudes of e-P3a and l-P3a (at Cz electrode site) is presented in Figure 5B.

Baseline-to-peak amplitudes (in microvolts) and peak latencies (in milliseconds, from the auditory stimulus onset to the maximum peak) of e-P3a and l-P3a were measured at Fz, Cz, and Pz electrode sites. Voltage and current source density (CSD) maps were obtained for topographic analysis (see Figure 7).

In order to evaluate the effects of the aMCI subtypes and the involuntary capture of attention provoked by novel stimuli, on the RT and the percentage of hits, two-factor ANOVAs with a between-subject factor Group (with three levels: Control, sdaMCI, and mdaMCI) and a within-subject factor Condition (with two levels: Standard, and Novel) were applied.

Two-factor ANOVAs (Group × Electrode Position) were applied, with the aim of evaluating the effects of sdaMCI and mdaMCI on e-P3a and l-P3a amplitudes. The Group factor included three levels (Control, sdaMCI, and mdaMCI) and the within-subject factor Electrode Position factor included three levels (Fz, Cz, and Pz).

One-factor ANOVAs (Group) were used to evaluate the effects of the aMCI subtypes on the e-P3a and l-P3a latencies at Cz, and this factor included three levels (Control, sdaMCI, and mdaMCI).

Pairwise comparison of means (with Bonferroni corrections) was carried out when the ANOVAs indicated significant results. In addition, partial eta squared (ηp2) and Cohen’s d value were calculated for each significant comparison, with the aim of determining the sizes of the effects. The Cohen’s d analysis was carried out with G^∗Power v.3.1.9.2 for Windows (Faul et al., 2009).

Receiver operating characteristics (ROC) curves were constructed when the Group factor exerted a main effect on the e-P3a and l-P3a parameters. As in our previous studies (Lindín et al., 2013), the ROC curve was constructed by determining the sensitivity and specificity for a range of values of the e-P3a and l-P3a parameters, and the tests were considered ideal when the area under the curve (AUC) was greater than 0.7. In addition, a line diagram was constructed with the sensitivity (true positive rate) plotted on the vertical axis and the false positive rate (1 minus specificity) on the horizontal axis.

Results were considered statistically significant at p ≤ 0.05. All statistical analyses were performed using the IBM SPSS Statistics package v.19 for Windows.

The ANOVA (Group × Condition) showed significant effects of the Condition factor [F(1,46) = 79.5; p < 0.001; ηp2 = 0.64] and the Group × Condition interaction [F(2,46) = 5.9; p = 0.005; ηp2 = 0.20] on the RT; this parameter was significantly longer (see Figure 2) in the Novel condition [mean (M) = 691 ms; standard deviation (SD) = 99.5] than in the Standard condition (M = 651 ms; SD = 82.7) in all three groups (post hoc comparison: Control: p < 0.001, d = 6.99; mdaMCI: p = 0.005, d = 7.31; sdaMCI: p < 0.001, d = 8.46).

The ANOVA (Group × Condition) showed a significant effect of the Group factor on the percentage of hits [F(1,46) = 8.7; p < 0.001; ηp2 = 0.27]; this parameter was significantly higher (see Figure 2) in the Control (M = 95.6%; SD = 4.1) and sdaMCI (M = 95.0%; SD = 5.4) groups than in the mdaMCI (M = 85.9%; SD = 11.4) group (post hoc comparisons of mdaMCI vs. sdaMCI: p = 0.002, d = 1.17; mdaMCI vs. Control: p = 0.001, d = 1.2; Control vs. sdaMCI: p = 1.00, d = 0).

For the e-P3a amplitude (see Figure 5A and Table 2), the ANOVA (Group × Electrode Position) revealed significant effects of the Group [F(2,47) = 3.4; p = 0.043; ηp2 = 0.13] and Electrode Position [F(2,94) = 33.8; p < 0.001; 𝜀 = 0.7; ηp2 = 0.42] factors, and for the Group × Electrode Position interaction [F(4,94) = 2.9; p = 0.026; ηp2 = 0.11]. The e-P3a amplitude was significantly larger in mdaMCI than sdaMCI at Pz (post hoc comparisons of mdaMCI vs. sdaMCI: p = 0.017, d = 0.98; mdaMCI vs. Control: p = 0.637, d = 0.50; Control vs. sdaMCI: p = 0.172, d = 0.64), and close to being significantly larger in mdaMCI than sdaMCI at Cz (p = 0.068, d = 0.83). In all groups, the e-P3a amplitude was also significantly larger at the Cz than at the Fz (p < 0.001) and Pz (p < 0.001) electrode sites, and in sdaMCI group it was significantly larger at the Fz than at the Pz electrode site (p = 0.009).

The ROC curves for e-P3a amplitude showed sensitivity and specificity values of 0.91 and 0.68, respectively (for more details see Table 3) for discriminating between the mdaMCI and sdaMCI groups at the Pz electrode site (positive group: mdaMCI, negative group: sdaMCI).

For the l-P3a amplitude (see Figure 5A and Table 2), the ANOVA (Group × Electrode Position) showed significant effects of the Group [F(2,47) = 6.3; p = 0.004; ηp2 = 0.21] and Electrode Position [F(2,94) = 34.0; p < 0.001; 𝜀 = 0.8; ηp2 = 0.42] factors, and for the Group × Electrode Position interaction [F(4,94) = 2.4; p = 0.055; ηp2 = 0.09].

The l-P3a amplitude was significantly larger in mdaMCI than in sdaMCI (p = 0.004) and Control (p = 0.017) groups at Fz (post hoc comparison of mdaMCI vs. sdaMCI: p = 0.035, d = 0.92; mdaMCI vs. Control: p = 0.032, d = 0.98; Control vs. sdaMCI: p = 1.00, d = 0.00), Cz (post hoc comparison of mdaMCI vs. sdaMCI: p = 0.006, d = 1.14; mdaMCI vs. Control: p = 0.014, d = 1.08; Control vs. sdaMCI: p = 1.00, d = 0.14) and Pz (post hoc comparison of mdaMCI vs. sdaMCI: p = 0.002, d = 1.28; mdaMCI vs. Control: p = 0.042, d = 0.99; Control vs. sdaMCI: p = 0.498, d = 0.47) electrode sites. Moreover, in all groups, the l-P3a amplitude was also significantly larger at the Cz than at the Fz (p < 0.001) and Pz (p < 0.001) electrode sites, and in Control group it was close to being significantly larger at Pz than at Fz (p = 0.060) electrode sites.

The ROC curves for l-P3a amplitude showed sensitivity and specificity values of, respectively, 0.73 and 0.79 at Fz, respectively, 0.64 and 0.74 at Cz, and, respectively, 0.82 and 0.68 at Pz for discriminating between mdaMCI and sdaMCI groups (positive group: mdaMCI, negative group: sdaMCI).

On the other hand, the ROC curves for l-P3a amplitude showed sensitivity and specificity values of, respectively, 0.73 and 0.75 at Fz, respectively, 0.73 and 0.60 at Cz, and, respectively, 0.64 and 0.70 at Pz for discriminating between mdaMCI and Control groups (positive group: mdaMCI, negative group: Control).

One-factor ANOVAs for the e-P3a and l-P3a latencies did not reveal a significant effect of the Group factor.

Voltage maps (see Figure 7) for e-P3a and l-P3a also showed larger amplitudes in the mdaMCI than in the sdaMCI and Control groups. CSD maps (see Figure 7) revealed different sources for e-P3a and l-P3a in all three groups: a centro-frontal source was observed for e-P3a, and frontal and parietal sources (with the parietal source increased in the mdaMCI) were observed for l-P3a. These results are consistent with previous research findings in healthy adults (see Correa-Jaraba et al., 2016). In addition, frontal (in control and sdaMCI participants) and occipital (in sdaMCI and mdaMCI participants) sinks for e-P3a, and temporal sinks (in control and mdaMCI participants) for l-P3a, were observed.

The mdaMCI participants performed worse than the Control and sdaMCI participants, as reflected in the lower percentage of hits in the first group. To our knowledge, only two previous studies have evaluated differences in performance between sdaMCI and mdaMCI participants using an A-V task (Cid-Fernández et al., 2017a,b). The authors observed longer RT and fewer hits in mdaMCI participants than in control (Cid-Fernández et al., 2017a,b) and sdaMCI (Cid-Fernández et al., 2017b) participants. In the present study, our results are partly consistent with the obtained by (Cid-Fernández et al., 2017a,b), as they revealed a behavioral decline in the mdaMCI participants relative to the sdaMCI and Control participants, as reflected in a lower percentage of hits, although there were no differences in RT between groups.

On the other hand, longer RTs in response to Go visual stimuli were observed in the Novel relative to the Standard condition, suggesting a distraction effect triggered by novel stimuli in all three groups under study. This finding is consistent with those of previous studies using A-V task in healthy young adults (Escera et al., 1998, 2001; Cid-Fernández et al., 2014, 2016; Correa-Jaraba et al., 2016) and middle-aged and old adults (Cid-Fernández et al., 2014, 2016; Correa-Jaraba et al., 2016).

In the three groups of participants, two successive phases of P3a were distinguished: an early phase (e-P3a) and a late phase (l-P3a). Both phases showed maximum amplitude at the Cz electrode site and were significantly larger at Cz relative to Fz and Pz electrode sites. In addition, a fronto-central distribution was observed for e-P3a, while l-P3a showed a parieto-central distribution, as observed in the voltage maps (Figure 7).

In a recent study with the same task as used in the present study, two phases of P3a were also identified in three different healthy age groups (Young, Middle-aged, and Old), in response to the novel auditory stimuli: e-P3a with a fronto-central distribution and l-P3a with a parieto-central distribution (Correa-Jaraba et al., 2016). In the aforementioned study, it was suggested that e-P3a may be a correlate of the orienting response to unexpected infrequent novel stimuli and that l-P3a may be a correlate of evaluation of these stimuli. This interpretation was based on the scalp distribution shown by the two P3a phases and on the findings of some previous studies (as cited in Correa-Jaraba et al., 2016). Specifically, differences were found for the frontal and parietal P3a (see Friedman et al., 2001), as P3a showed higher habituation at frontal than at parietal sites (Courchesne et al., 1975; Knight, 1984; Friedman and Simpson, 1994). The frontal part of P3a was considered an index of processes associated with orienting toward the stimulus (Cycowicz and Friedman, 1997; Friedman et al., 1998), because habituation is a distinctive characteristic of the orienting response (Siddle, 1991; Öhman, 1992); on the other hand, the posterior part of P3a was interpreted as probably reflecting categorization processes (Courchesne et al., 1975; Knight et al., 1995), because it showed similar features to the P3b component evoked in response to target stimuli (Friedman et al., 2001).

In the present study, both the distribution (central and parietal) and latency of l-P3a were similar to those of the P3b component. We therefore suggest that these ERP components may be identical, although further studies should be carried out to explore this possibility.

On the other hand, the findings of the present study are only partly consistent with those reported by Escera et al. (1998, 2001), who using a similar task observed two P3a consecutive phases, in response to novel auditory stimuli, in young participants: (1) an early phase, peaking between the 220 and 320 ms and with a central distribution, and (2) a late phase, with latencies about 300–400 ms and a right frontal scalp maximum. The authors proposed that the late P3a phase may reflect orienting of attention toward the irrelevant novel stimulus, because its amplitude increased when these stimuli could act as warning signals for consecutive relevant visual stimuli, relative to conditions where they did not act in this way (in a passive oddball task). On the other hand, the early P3a phase was interpreted as an index of a violation (provoked by a novel stimulus) of the regularity of an established environment model (as cited in Correa-Jaraba et al., 2016). The authors suggested that the early P3a phase may not reflect orientation toward the stimulation because “the amplitude did not increase when it acted as a warning signal, unlike in the pure passive oddball task” (as cited in Correa-Jaraba et al., 2016).

Our findings revealed significant differences between the three groups of participants for the amplitude of e-P3a and l-P3a phases; however, they did not show differences between groups for the latencies.

The e-P3a amplitude was significantly larger in adults with mdaMCI than in adults with sdaMCI, which may indicate a greater involuntary capture of the attention toward the irrelevant novel stimuli in the first. Moreover, the increase in the e-P3a amplitude seems be a potential biomarker for discriminating between mdaMCI from sdaMCI participants, with good sensitivity and acceptable specificity (see Table 3).

The l-P3a amplitude was also significantly larger in the mdaMCI groups than in sdaMCI and Control groups, which may indicate that participants with mdaMCI allocate more attentional resources for evaluating irrelevant novel stimuli. The increase in the l-P3a amplitude also seems to be a potential biomarker for mdaMCI, as it was able to discriminate between the mdaMCI and Control groups, and mdaMCI and sdaMCI groups, with moderate sensitivity and specificity (see Table 3).

Further studies should be carried out with larger samples of participants to confirm our results and also to evaluate the clinical usefulness of changes in e-P3a and l-P3a amplitudes for identifying aMCI subtypes.

The observed differences between the mdaMCI group and the other two groups (healthy controls and sdaMCI), together with the absence of significant differences between the sdaMCI and Control groups, provide evidence for the need to distinguish aMCI subtypes with the aim of evaluating the P3a component as a neurocognitive marker for diagnosing aMCI. Thus, it is possible that the similarity between the P3a amplitudes of sdaMCI and Controls may mask the differences between mdaMCI and Controls.

The need to distinguish the aMCI subtypes was also evidenced in our previous ERP studies in which other (Cespón et al., 2015) or similar (Cid-Fernández et al., 2017a,b) cognitive tasks were used. These studies revealed behavioral and neurocognitive decline in mdaMCI participants relative to the sdaMCI and control participants. Specifically, a decrease in the allocation of attentional resources to target stimuli was observed in mdaMCI participants relative to healthy controls, with no difference between controls and sdaMCI participants (Cespón et al., 2015), as well as deficits in executive processes in mdaMCI relative to sdaMCI and control adults (Cespón et al., 2015; Cid-Fernández et al., 2017a,b). Studies using neuroimaging techniques or histopathologic analysis also supported the need to differentiate the two aMCI subtypes, as adults with mdaMCI showed different brain damage than those with sdaMCI, e.g., a more widespread white matter degeneration (Li et al., 2013; Liu et al., 2017) and higher β-amyloid (Aβ) deposition (for a review see Ong et al., 2013).

Several follow-up studies showed that the risk of conversion to AD is higher in mdaMCI than in sdaMCI (Tabert et al., 2006; Caffarra et al., 2008). In the present study, the larger e-P3a and l-P3a amplitudes shown by the mdaMCI group than by the sdaMCI group may also be optimal markers of possible progression to AD; however, this assumption would be better addressed in a longitudinal study comparing aMCI subtypes.

Cecchi et al. (2015) obtained smaller P3a amplitude in adults with mild AD than in healthy controls; however, other authors found no difference in this parameter between both groups (Yamaguchi et al., 2000). Given that patients with AD showed changes in P3a parameters relative to controls, we expected to find similar changes in aMCI, especially in the mdaMCI group. In this respect, the results of the present study are partly consistent with the findings for AD, because we observed significant differences between the mdaMCI group and the sdaMCI and/or Control groups, with significantly larger e-P3a and l-P3a amplitudes in the mdaMCI group. In addition, we did not observe any differences between the sdaMCI and Control groups in e-P3a or l-P3a parameters. First, it should be noted that reports of the effects of AD, and other types of dementia, on P3a are scarce, and the findings are to some extent inconsistent. Moreover, due to the lack of research on the P3a component in people with aMCI, comparison of our findings with other relative findings is difficult. The discrepancies between the results of the present and previous studies with AD patients may stem from variations in the methods used, e.g., related to different task characteristics.

In summary, healthy, sdaMCI, and mdaMCI participants performed an auditory-visual distraction-attention task while EEG activity was recorded. A greater orienting response to unattended novel auditory stimuli, and allocation of more attentional resources for the subsequent evaluation of these stimuli (as indicated by larger e-P3a and l-P3a amplitudes, respectively) were observed in the mdaMCI group relative to the sdaMCI and Control groups. In addition, the sensitivity and specificity scores provide evidence supporting the use of larger e-P3a and l-P3a amplitudes as neurocognitive markers of mdaMCI in the clinical setting. Follow-up studies should investigate the predictive utility of e-P3a and l-P3a amplitudes to identify people with aMCI who will develop AD and those who will remain stable.