Intensity-dependence of auditory-evoked potentials might present an early surrogate marker for post-stroke depression

(cid:1) Intensity-dependence of auditory-evoked potentials present early surrogate marker for post-stroke depression (cid:1) Early intensity-dependence of auditory-evoked potentials increased in later depressed stroke patients (cid:1) Early intensity-dependence of auditory-evoked potentials correlates with depressive symptoms of all stroke patients


Introduction
Post-stroke depression (PSD) is a frequent complication following stroke and has devastating impact on both stroke outcome and recovery.It is defined as the occurrence of depressive symptoms following stroke after an interval of at least 2 weeks.Although often underdiagnosed -and, more importantly, undertreatedapproximately one third of stroke survivors suffer from PSD (Robinson, 2006;Ayerbe et al., 2013;Hackett and Pickles, 2014).
PSD impedes stroke recovery in various ways: it impairs cognition, influences motor functioning and even increases mortality (Johnson et al., 2006;Robinson, 2006;Bartoli et al., 2013;Ayerbe et al., 2014).The strongest relationship exists between PSD and activities of daily living, which are correlated reciprocally (Johnson et al., 2006).
Although the exact causes for PSD remain unknown, various pathophysiological models have been introduced to explain the biological mechanisms resulting in PSD (Robinson and Bloom, 1977;Spalletta et al., 2006;De Ryck et al., 2014;Li W et al., 2014).Ultimately, all pathophysiological processes result in central-nervous serotonin deficiency which is known to play a major role in the development of depression (Heninger et al., 1996).
This serotonergic deficiency can be detected by measuring the intensity-dependence of auditory-evoked potentials (IDAP) (Hegerl and Juckel, 1993).An increased IDAP indicates low central serotonergic tone (Juckel et al., 1999).It therefore serves as a noninvasive marker of behavior-relevant serotonin level (Hegerl and Juckel, 1993;Juckel et al., 1999).Increased IDAP has been found in patients suffering from various psychiatric conditions such as depression or borderline personality disorder (Norra et al., 2003;Linka et al., 2009).Different studies also showed that IDAP can help detect depressed patients that are more likely to respond to antidepressant-treatment (Paige et al., 1994;Gallinat et al., 2000;Linka et al., 2004): Gallinat et al. demonstrated that depressed patients that showed an increased IDAP prior to treatment with selective serotonin reuptake inhibitors (SSRI) had higher response rates to treatment than those who did not (Gallinat et al., 2000).Rocco et al. found that, in the acute stroke setting, IDAP was significantly increased in patients suffering from subcortical stroke compared with healthy controls (Rocco et al., 2007).They also observed a positive correlation between IDAP and the severity of depressive symptoms in the group of acute stroke patients.These findings suggest an early involvement of the central serotonergic system following acute stroke.
Various potential surrogate markers for PSD have been investigated, including imaging markers (e.g.lesion location), molecular markers (e.g.brain-derived neurotrophic factor (BDNF)), genetic markers (e.g.polymorphisms in serotonin transporter gene SLC6A4) and perceived stress following stroke (Kohen et al., 2008;Laures-Gore and DeFife, 2013;Douven et al., 2017;Levada and Troyan, 2018).Although some of these markers seem promising, there still is a lot of conflicting evidence in the literature and many have yet to be assessed for their predictive potential.However, stroke survivors would greatly benefit from markers that can early identify PSD risk enabling preventive measures.
Despite the high PSD incidence and its compromising impact on stroke recovery, a predictive surrogate marker for the development of PSD remains to be clinically established.Aim of this study was to evaluate whether increased IDAP following a stroke can predict the development of PSD and thus serve as an early surrogate marker.We therefore hypothesized a possible correlation between IDAP early after and the occurrence of PSD from 4 weeks after stroke.

Patient recruitment
All patients admitted to the stroke unit of Charité-Universitäts medizin Berlin, Campus Benjamin Franklin were screened by a hospital-intern Trial Team for study participation.Eligible were all patients over the age of 18 years with a first-ever acute MRIproven ischemic stroke.To rule out pharmacogenic confounders, we excluded patients taking beta blockers, antiseizure medication, antidepressants or triptanes.Additionally, patients having a stroke localized in thalamus or temporal cortex were excluded from study participation since these regions are involved in the AEP generation (Koella, 1985;Mäkelä and Hari, 1990;Hegerl and Juckel, 1993;Bagdy et al., 2007;Peterlin and Rapoport, 2007).Table 1 shows the entire exclusion criteria.

Study protocol
Participants underwent clinical assessments and measurement of auditory-evoked potentials (AEP) at 2 different time points.The first measurement was conducted 24 to 72 h after stroke onset.The second assessment took place not earlier than 30 days, but latest 6 months after stroke either in the hospital or, if not feasible, via telephone call.In the latter case, we performed a standardized phone questionnaire, while AEPs were not obtained.Written consent of all patients or their legal guardians were obtained after detailed information and medical education.The study protocol was approved by the Ethics Committee of the Charité-University Medicine Berlin.

Clinical assessments
On admission, all patients received a brain MRI.We assessed every patient on both measurement points for stroke severity using National Institute of Health Stroke Scale (NIHSS) score, for depressive symptoms using Montgomery-Åsberg Depression Rating Scale (MADRS) score and for impediments in activities of daily living using Barthel index and Modified Rankin Scale (MRS) score.In addition, we thoroughly clinically examined and questioned every patient concerning their medical and medication history.

AEP acquisition and IDAP calculation
The recording of AEPs took place in a calm room in a comfortable chair.In each subject, we determined monaural hearing threshold levels, consecutively from both ears, with passively noise cancelling headphones before AEP recordings to ensure comparability of our results.Auditory stimuli with intensities of 60, 70, 80 and 90 dB sensation level (i.e., above individual hearing thresholds) were presented binaurally in 4 randomized blocks of 90 trials each.The stimuli consisted of 1000 Hz tones (total length 40 ms, each 10 ms rise and fall, audio sampling rate 44100 Hz), with random interstimulus intervals in the range of 1000 ± 200 ms.
EEG was recorded at a sampling rate of 5000 Hz with electrodes at the left mastoid, referenced to C1, and at the right mastoid, referenced to C2, with ground electrode at AFz according to the 10-10  (Hegerl and Juckel, 1993;Rocco et al., 2007). 2 Listed medications influence the central serotonergic system (Koella, 1985;Mäkelä and Hari, 1990;Peterlin and Rapoport, 2007).
N. Meißner-Bendzko, G. Waterstraat, G. Curio et al. Clinical Neurophysiology 167 (2024) 254-261 system.The standard recommendation for AEP recording is a 2channel lead configuration with mastoids (ipsi/contralateral) against Cz (Nuwer et al., 1994).We chose to use electrodelocations C1 and C2 instead of Cz for our set up because this allowed for a more precise analysis of AEP lateralization while only deviating from Cz by 10% (approx.3.5 cm) of the spherical distance between both preauricular points.Electrode impedances were kept below 5 kX.Hardware filters comprised a 0.1 Hz high-pass and a 1000 Hz antialiasing low-pass filter.Brain Vision Analyzer 2.2 (Brain Products, Munich, Germany) was used for offline EEG analysis.EEG was digitally filtered at 1 Hz high-pass and 20 Hz low-pass with 24 dB/octave each and re-referenced to common average.Artifacts (! 50 lV at any electrode) were marked and removed (200 ms before and after the artifact).The EEG was then segmented according to the conditions of 60, 70, 80 and 90 dB, followed by a baseline demeaning in the range from À 50 ms to 0 ms.We removed the first five AEPs of each condition to eliminate adaptation effects; the remaining segments were then averaged for each condition.Averaged AEP segments consisted of periods from 50 ms before to 400 ms after stimulus onset.For further analysis, AEP of the left and right hemispheres were assessed both separately and in a pooled analysis.If not indicated otherwise, results are reported from the pooled analysis.The N1 component was determined baseline-to-peak as the maximal negative deflection in a latency range from 50 to 150 ms after stimulus onset; the P2 component was determined baseline-to-peak as the largest positive deflection in a latency range from 100 ms to 250 ms after stimulus onset.
We calculated the N1-P2 amplitude of each sound intensity as the difference between mean N1 and P2 peaks.By means of linear regression of these mean amplitudes, we obtained the linear amplitude/stimulus intensity function (ASF) (l V/10 dB), defined as IDAP.

Statistical analysis
We performed statistics using SPSS 29.0 (IBM, Armonk, US).We compared differences in ASF slopes between groups with Student's t-test.Correlation analysis was done with Spearman rho correlation test and regression analysis was performed with univariate analysis of variance.Level for statistical significance was set at p < 0.05 for all tests.Results are, if not indicated otherwise, reported as mean ± standard deviation.

Participants included
During the 24-month recruitment period, we included 66 S patients.39 patients finished the study; 32 had in-hospital follow-up and 7 telephone follow-up.Two patients who had inhospital follow-up needed to be excluded from the final analysis because of considerable artifacts in the AEP-recordings.We thus carried out the final analysis with data of 37 patients.12 participants were female, 14 had left-sided stroke.The mean NIHSS score at admission was 1.43 ± 1.82.Table 2 shows detailed patient data.
A MADRS score !7 at follow-up was used to define PSD (Montgomery and Åsberg, 1979;Müller et al., 2003;Spalletta and Robinson, 2010).We divided the study participants in 2 groups: those with depressive symptoms (MADRS !7 points) at follow-up and those without (MADRS < 7 points).Fig. 1 shows the averaged AEP curves of all participants at first measurement.

Occurrence of PSD
We diagnosed PSD in 7 out of 37 patients (19%) during followup (mean MADRS score 14.29 ± 6.65).15 patients from the nondepressed group (n = 30) also reported depressive symptoms that did not reach the threshold of MADRS ! 7.
At the first measurement point, we did not observe a correlation between IDAP and MADRS scores (r = 0.25; p = 0.137).

Correlation of IDAP on admission and MADRS on follow-up
We compared the groups of depressed and non-depressed patients -based on MADRS-scores obtained at follow-up -with regard to IDAP assessed in the acute stage after the stroke.The PSD group showed significantly steeper ASF slopes compared with the non-depressed group (0.76 ± 1.14l V/10 dB vs. À 0.09 ± 0.58l V/10 dB; p = 0.007).Fig. 2 demonstrates the ASF slopes of both groups on admission.There were no statistically significant differences regarding stroke laterality (right/left/bilateral), age, sex, stroke severity or disability in the acute stroke phase (see Table 2).For all stroke patients as a group, we also found a positive correlation between early assessed ASF slopes and subsequent MADRS-scores (r = 0.44; p = 0.007).Fig. 3 demonstrates the corresponding scatterplot.A univariate analysis with dependent variable MADRS-scores on follow-up, fixed variable sex and covariates early assessed ASF slopes and age showed a coefficient of determination of 0.26 (p = 0.018) and a regression coefficient for pooled ASF slopes of 3.77 (p = 0.004).The variables sex and age were non-significant.
Interestingly, these findings were even more pronounced when we analyzed AEP recordings without pooling over hemispheres but separately with regard to the lesion site: For recordings ipsilateral to the lesion site, the ASF slopes on first measurement point were significantly higher in later depressed than in non-depressed patients (1.31 ± 0.74l V/10 dB vs. À 0.51 ± 0.66l V/10 dB; p = 0.0003).For contralateral recordings, the differences were still statistically significant but less pronounced (1.04 ± 0.74l V/10 dB vs. 0.10 ± 0.61l V/10 dB; p = 0.006).Also, the positive correlation between early ASF slopes and later MADRS-scores in all stroke patients as a group was stronger in ipsilateral AEP recordings compared with contralateral recordings (r = 0.69; p = 0.00003 vs. r = 0.48; p = 0.009).For these lesion-site specific analyses, we excluded 7 patients suffering from bilateral stroke.
Additionally, we found that the correlation between ipsi À and contralesional ASF slopes on admission was less pronounced than on follow-up (r = 0.55 on admission vs. r = 0.72 on follow-up).

Correlation of IDAP and MADRS on follow-up
There was no statistically significant correlation between IDAP and MADRS scores both assessed on last measurement point (r = 0.06; p = 0.758).The difference regarding ASF slopes assessed on last measurement point between depressed and non-depressed patients did not reach statistical significance, although there was an observable trend (0.71 ± 0.91l V/10 dB vs. 0.27 ± 0.71l V/10 dB; p = 0.24).
For this follow-up analysis, we excluded 7 patients for which we did not obtain further AEPs because they only had telephone follow-up.We compared the telephone follow-up group (n = 7) with the in-hospital follow-up group (n = 30) regarding MADRS scores at follow-up and mean ASF slopes on admission: Both the mean MADRS score (7.14 ± 9.63 vs. 3.00 ± 4.81; p = 0.104) and the mean ASF slope (0.50 ± 0.97l V/10 dB vs. À 0.26 ± 0:70 l V/10 dB; p = 0.109) were higher in the telephone group, although these differences were statistically non-significant.
The groups of depressed and non-depressed patients differed significantly regarding MRS scores obtained on follow-up (1.43 ± 0.98 vs. 0.53 ± 0.68; p = 0.007) and for all patients there was a trend towards a positive correlation between MADRS scores and MRS scores on follow-up (r = 0.32; p = 0.054).Our analysis also revealed a non-significant difference in NIHSS scores assessed on follow-up between the two groups (1.57± 1.72 vs. 0.7 ± 0.99; p = 0.08).

Time span between measurements as possible confounder
The timing of follow-up ranged from 30 to 192 days after first measurement with a mean of 83.3 days and a median of 81 days.To assess for a possible confounding effect of follow-up timing, we performed correlation analyses to evaluate the relationship between the time span between measurement time points and our outcome measures at follow-up, including MADRS, NIHSS, MRS and Barthel index.We further compared the groups of later depressed and non-depressed patients regarding the timing of follow-up, using an independent sample t-test.
The analyses revealed no statistically significant correlations between the time span and the different measures and no statistically significant differences in follow-up timing between the depressed and non-depressed patients (79.3 ± 29.9 days vs. 84.2± 41.5 days; p = 0.768).

Dropouts
27 of 66 patients (41%) did not complete the study.The average ASF slope of dropouts on first measurement was 0.26 ± 0.92l V/10 dB and the average MADRS score was 3.82 ± 5.98.

Key learnings
Our study found a significant difference in ASF slopes assessed early after stroke between depressed and non-depressed patients at follow-up.We also found a positive correlation between early ASF slopes and subsequent MADRS-scores for all stroke patients as a group.Both findings support the hypothesis that the IDAP can predict the development of depressive symptoms and therefore serve as an early surrogate marker for PSD.To our knowledge, this is the first ever longitudinal study to assess the relationship between IDAP and PSD.

PSD pathophysiology and markers
The data also show that the positive correlation between early ASF slopes and depressive symptoms on follow-up is even more pronounced in ipsilesional recordings compared with contralesional or pooled recordings.This finding is consistent with different pathophysiological concepts on the development of PSD, focusing on central serotonin deficiency due to neuroinflammation: First, ischemic brain lesions lead to localized activation of microglia which release several pro-inflammatory cytokines (Spalletta et al., 2006).These can induce the enzyme indolamine 2,3-dioxygenase (Spalletta et al., 2006;Li W et al., 2014).Second, those cytokines also amplify localized inflammatory processes which enhance apoptosis of limbic neurons (Spalletta et al., 2006).Third, stroke induces localized apoptosis of biogenic amincontaining neurons projecting from the brainstem to the cortex (Robinson and Bloom, 1977).All three pathophysiological processes alter central serotonin metabolism and reduce hippocampal neurogenesis and synaptic plasticity which contribute to the development of depression (Sandu et al., 2015).
The reduced serotonergic activity can be measured via AEPs.The AEP assessment right after stroke might have detected those localized pathological processes which could explain why IDAP was initially more pronounced when measured ipsilateral to stroke site (Robinson et al., 1977).The correlation between ipsi-and contralesional ASF slopes on admission was less pronounced than on follow-up (r = 0.55 vs. r = 0.72).This finding supports the assumption that differences in early assessed AEPs are a result of early localized changes in serotonin level, while the smaller differences between AEPs assessed later are a result of overall serotonin deficiency, which is no longer lesion-site specific.
In fact, many factors implicated in PSD's etiological pathways have been evaluated and proposed as potential surrogate markers.Particularly neuroimaging markers have been subject of ongoing debate (Carson et al., 2000;Robinson and Jorge, 2016;Douven et al., 2017).During the first year following stroke, lesions involving the basal ganglia, thalamus and prefrontal cortex are associated with an increased risk for PSD (Douven et al., 2017;Levada and Troyan, 2018).The relationship between left hemispheric stroke and PSD remains inconclusive and it seems like this correlation may be time dependent (Carson et al., 2000;Levada and Troyan, 2018;Robinson and Jorge, 2016).Considerable progress has been made in the identification of blood markers for PSD.Particularly noteworthy are the growth factors IGF-1 and BDNF, as their respective peripheral increase or decrease early after stroke was linked to the development of PSD, suggesting their involvement in the pathophysiological processes leading to PSD (Li J et al., 2014;Levada and Troyan, 2018;Xu et al., 2018).A genetic predisposition seems to exist for PSD as certain gene polymorphism have been found that are associated with depressive symptoms following stroke.These include the 5-HTTLPR and the STin2 VNTR polymorphisms of the serotonin transporter gene SLC6A4 as well as hypermethylation of the BDNF gene (Kohen et al., 2008;Kim et al., 2013).Elevated levels of various serum pro-inflammatory cytokines, including IL-1b, IL-6, IL-18, TNF-a and IFN-c have been consistently linked to PSD, providing strong evidence that not only neuroinflammation but also systemic inflammatory processes contribute to PSD development (Spalletta et al., 2006;Levada and Troyan, 2018;Robinson and Jorge, 2016).Other markers of inflammation include ferritin, neopterin, glutamate and C-reactive protein (Levada and Troyan, 2018).HPA axis dysregulation, which shows a bidirectional relationship with pro-inflammatory cytokines, seems to play a role since as well since hypercortisolemia and reduced negative cortisol feedback have been found in depressed stroke survivors (Robinson and Jorge, 2016;Levada and Troyan, 2018).
It is worth also discussing other electrophysiological markers, such as sleep-like slow waves following therapeutic focal cortical lesions (Russo et al., 2021).Interestingly, these slow waves extend beyond the peri-lesional region, implicating broader network-level effects.Previous studies have linked sleep-like slow waves to the serotonin system and cognitive impairment (Jouvet 1999;Vyazovskiy et al., 2011;Russo et al., 2021).Thus, IDAP and slow wave dynamics share some interesting commonalities: they both correlate with alterations in the central serotonergic system, are associated with cognitive impairment and reflect larger scale network effects of focal lesions.Determining whether both phenomena arise from serotonin alterations or contribute to them remains to be further investigated.
Beyond biological factors, several psycho-social factors have been suggested as PSD markers, including both physical and cognitive impairment, stroke severity and social factors such as social isolation and living alone (Hackett and Anderson, 2005).Perceived stress also strongly correlates with depressive symptoms poststroke (Laures-Gore and DeFife, 2013).
While some of the identified biological and psycho-social markers seem promising, it is important to note that most of them have not been investigated in prospective studies and primarily only after depressive symptoms already had developed.Their ability to reliably detect PSD risk remains to be investigated and therefore, these markers cannot yet be recommended for routine clinical screening for PSD.In our study, for example, MRS scores as measure for disability following stroke correlate with later depressive symptoms as described by other authors (Hackett and Anderson, 2005).However, this correlation only exists for MRS scores obtained at follow-up but not in the acute stroke phase, meaning it has no predictive value in our study.In contrast, IDAPs advantage lies in its predictive potential early after stroke, making it a compelling candidate for an early surrogate marker for PSD.

IDAP and MADRS correlation when assessed at the same time
IDAP is known to indicate central serotonergic activity and thus to correlate with depressive symptoms (Hegerl and Juckel, 1993;Juckel et al., 1999).We used this well-explored finding as the basis for our hypothesis and indeed demonstrated a correlation between early IDAP and the development of depressive symptoms.
Our data, however, did not show a correlation between depressive symptoms and ASF slopes when observed at the same time which one would expect.
Concerning the comparison on the last measurement point, there was a trend towards a positive correlation.Of 37 patients, 7 (19%) only had telephone follow-up and their AEP data was thus missing for the follow-up analysis.While the comparison between telephone-based and in-hospital follow-up patients indicated a steeper ASF slope at admission and higher MADRS scores at follow-up for the telephone group, the small sample size does not allow any definitive conclusions.The correlation between depressive symptoms and ASF slopes at follow-up for all stroke patients therefore requires validation using a larger study cohort.
A hypothesis for depressive symptoms and IDAP on first measurement point not correlating is that ischemic stroke lesions lead to imbalances in the serotonergic system that can be immediately measured by ASF slopes of AEPs, but the clinical consequences À meaning development of PSD À need time to develop.This is also reflected by the definition of PSD which states that depressive symptoms occur earliest 2 weeks after stroke.In fact, this time gap between causal brain lesion resulting in serotonin imbalance and clinical presentation of PSD is the fundament of our hypothesis that IDAP can indicate early changes in serotonergic tone and can therefore predict the development of PSD.

Limitations
The relatively small number of enrolled participants mainly resulted from the strict inclusion and exclusion criteria.Especially the intake of ß-blockers was a limiting factor since they are commonly used in stroke patients.
The observed PSD incidence of 19% is lower than the expected incidence of approximately one third given in the literature (Ayerbe et al., 2013;Hackett and Pickles, 2014).However, most of the recent meta-analyses on PSD incidence examined the development of PSD in a wide period ranging from a few days up to 5 years after stroke (Ayerbe et al., 2013;Hackett and Pickles, 2014).This lower percentage might also be explained by a selection bias: Patients who suffered from sadness and lethargy immediately after stroke were more likely to deny participation.We thus included mostly patients who suffered only little from stroke sequelae.The same bias is possible when looking at stroke severity: The average NIHSS score was 1.43 ± 1.82, which represents a minor stroke (Yakhkind et al., 2016).
We investigated the timing of follow-up as a possible confounder because the timing ranged from 30 to 192 days, largely because participants were engaged in ongoing rehabilitation and therefore frequently not available for follow-up.Even if the follow-up assessment was performed both for the time of the interview and retrospectively, it is possible that depressive symptoms were missed due to this circumstance.Our analyses, however, did not reveal any significant correlations between the follow-up timing and any of the assessed measures such as MADRS scores and the later depressed and non-depressed patients did not differ significantly in terms of time span between measurements.These findings suggest that the rather large span between the measurement time points does not represent a confounder.Although the time span seems not to affect the study results, it is well known that the timing of assessing depressive symptoms is crucial and PSD incidence given in the literature largely varies depending on the timing (Ayerbe et al., 2013;Hackett and Pickles, 2014).We therefore suggest that follow-up studies aim to narrow down the time window of follow-up despite the difficult circumstances such as rehabilitation which is such an important pillar of post-stroke recovery.
Only 59% of initially enrolled patients completed the study.The reasons cited for not completing the study, ranked by frequency, were: Having no time due to ongoing rehabilitation, having lost interest in the study and perceiving the AEP measurement as unpleasant.Another reason for the high dropout rate could, in fact, be the development of PSD in dropouts since some patients reported depressive symptoms when inquired about why they did not complete the study.This assumption is supported by the finding that some of those dropout patients showed considerably steep ASF slopes on first measurement point.
Our data show relatively high levels of standard deviation concerning ASF slopes (see Table 2).Although ASF slopes are intraindividual very stable and inversely correlate with serotonergic activity, Hegerl and Juckel (Hegerl and Juckel, 1993) demonstrated that ASF slopes show a high interindividual variability which explains the high standard deviations we observed.

Outlook
Although different markers have been proposed (Hackett and Anderson, 2005;Laures-Gore and DeFife, 2013;Li J et al., 2014;Douven et al., 2017;Levada and Troyan, 2018) there is currently no clinically well-established method to assess the risk of stroke survivors for PSD.Given the high PSD incidence and its devastating impact on stroke recovery, the clinical establishment of a predictive surrogate marker for PSD would be of great use to stroke patients.
Also, early detection could imply therapeutic consequences: not only did studies show that the use of SSRI can attenuate depressive symptoms (Allida et al., 2020), but also that early treatment may prevent PSD development (Robinson et al., 2008;Salter et al., 2013).Rat models showed that administration of antidepressants can even reverse the biological changes following a stroke that lead to depression (Wang et al., 2010;Kronenberg et al., 2012).Even in non-depressed patients, administration of SSRI after stroke may improve outcome regarding activities of daily living, motor recovery, and mortality (Chemerinski et al., 2001;Jorge et al., 2003;Chollet et al., 2011;Mikami et al., 2011).
Despite the promising data, we were not able to define an IDAPthreshold predicting PSD due to the relatively small sample size.We believe that further research with larger study size can help to find such thresholds and make IDAP a useful, non-invasive tool to identify stroke patients at higher risk for PSD À who also might benefit from early SSRI treatment.

Conclusion
Altogether, our data suggest that the IDAP might be used as early surrogate marker for PSD.

Fig. 1 .
Fig. 1.Averaged AEP curves of all patients on first measurement point.Averaged AEP (auditory-evoked potential) curves of all patients on first measurement, each for intensities 60, 70, 80 and 90 dB.N1: maximum negative deflection from 50 to 150 ms.P2: maximum positive deflection from 100 ms to 250 ms.

Fig. 2 .
Fig. 2. ASF slopes on admission of later depressed and non-depressed patients.Amplitude/stimulus intensity function (ASF) slopes on admission of depressed and nondepressed patients À categorized based on Montgomery-Åsberg Depression Rating Scale (MADRS) scores on follow-up.Depressed patients show significantly steeper ASF slopes on admission than non-depressed patients (p = 0.007).

Fig. 3 .
Fig. 3. Correlation between ASF slopes on admission and MADRS scores on follow-up.Scatterplot showing spearman rho correlation between pooled amplitude/stimulus intensity function (ASF) slopes on admission and Montgomery-Åsberg Depression Rating Scale (MADRS) scores on follow-up for all stroke patients (r = 0.44; p = 0.007).

Table 1
Study exclusion criteria.

Table 2
Demographics and clinical data of later depressed and non-depressed patients.