2.1.

Ethics Statement

All participants gave written informed consent before study inclusion. All procedures were performed in accordance with the Declaration of Helsinki and were approved by the Korea University Medical Center Institutional Review Board (IRB No. ED15363).

2.2.

Subjects

Patients with symptoms of OI were recruited from January 2016 to June 2016. For study accuracy, we rigorously adjusted the inclusion criteria. Inclusion criteria were (1) symptoms suggesting OI (memory loss, visual difficulties, light headedness, headache, fatigue, increases or decreases in BP, weakness, nausea and abdominal pain, sweating, tremulousness, and exercise intolerance),¹⁹ (2) no history of central nervous system diseases, including stroke, Parkinson’s disease, and Alzheimer’s dementia, and (3) no history of significant head injury, alcohol, or psychotropic drug abuse. Demographic and clinical data, including age, gender, and comorbid chronic diseases, were obtained. Normal healthy control (HC) volunteers who did not have any medical history that might affect ANS were recruited. Healthy subjects who had any abnormal findings in the AFTs of this study protocol and had any OI symptoms according to the OI questionnaires were excluded from final analysis. In addition, subjects with arrhythmia were excluded because of the possibility of VM measurement error. A total of 31 patients with OI and 9 healthy subjects who completed all study protocols were recruited. However, four patients and three healthy subjects who showed arrhythmia, which made parameters unreliable, were excluded from final analysis. Two patients who had NIRS signals that were not sufficient to be analyzed due to excessive motion artifacts were excluded. Another three healthy subjects were excluded due to abnormal VM (AbVM) parameter findings: two subjects showed abnormal Valsalva ratios (VRs) and one subject showed abnormal cardiovagal barosensitivity (BRSv). Finally, data from 25 patients and 3 normal HCs were analyzed.

2.3.

Study Design

All participants taking medications affecting autonomic function were asked to discontinue their medicine for at least 24 h prior to VM. To minimize the effects of confounds, VM was controlled and regulated by the standard electrodiagnostic laboratory environment.²⁰ Tests were performed in the following sequence: (1) autonomic dysfunction self-questionnaires were completed by the patient and (2) VMs were conducted while attaching the NIRS probe on the forehead to investigate cerebral hemodynamics.

2.4.

Autonomic Dysfunction Questionnaires

Self-report questionnaires were administered. Autonomic dysfunction was measured using the Korean version of the Composite Autonomic Symptom Score 31 (K-COMPASS 31) consisting of 31 items that assess the following categories: 4 items for OI, 3 items for vasomotor, 4 items for secretomotor, 12 items for gastrointestinal, 3 items for bladder, and 5 items for pupillomotor. The sum of items from each category and the sum of all 31 items were used for analysis. Of note, higher scores indicate more severe autonomic dysfunction.²¹ We also only scored the four items related to OI separately among the many items of the K-COMPASS 31 questionnaire.

2.5.

Valsalva Maneuver

The VM was performed with Finometer equipment (Finapres Medical Systems BV, Amsterdam, The Netherlands) with a cuff placed on the middle finger. Subjects blew through a mouthpiece attached to a manometer and maintained a pressure of 40 mm Hg for 15 s at rest and in the recumbent position. VM is a difficult procedure to perform consecutively, so each trial was followed by a rest period of 2 min, with testing repeated until two responses with similar beat-to-beat BP and HR recordings were obtained. The VM trial that showed the correct four phases of VM in the BP recordings was chosen for further NIRS analysis. Therefore, there was only one VM trial per subject in which BP and NIRS data were used for analysis. The following BP metrics were determined for each patient. (1) VR was calculated as the maximum HR during the maneuver in phase II divided by the lowest HR in phase IV. The VR of each subject was compared with age-specific reference values.^2,22 (2) Pressure recovery time (PRT) was defined as the time from the lowest systolic BP (SBP) during the valley of phase III (start of phase IV) to the return of SBP to baseline before overshooting of SBP during phase IV.^1,23 (3) Alternative adrenergic barosensitivity (${\mathrm{BRSa}}_1$) was defined as the sum of the SBP decrement during early phase II and two-thirds of the SBP decrement in phase III divided by PRT.^24,25 (4) BRSv was determined as the regression slope of the $R-R$ interval (milliseconds) over SBP during early phase II.^24–27 Using the cutoff value for each parameter, patients with an abnormal result in any of the four parameters were placed into the AbVM group.

2.6.

Near-Infrared Spectroscopy

A custom-built 108-channel NIRS system was administered to the subject pool to monitor cerebral hemodynamics during VM (Fig. 1). The NIRS probe was built to span the subject’s entire forehead and was secured in place by a medical band. The center bottom of the NIRS probe was aligned with the FpZ point, and the left and right corners were aligned with the F7 and F8 points, according to the International 10–10 system. This alignment has been shown to be appropriate for monitoring cerebral oxygenation changes in the prefrontal area.^28,29 The experiment administrator visually ensured good contact for all sources and detectors before the experiment started. In addition, the NIRS software included a module that identified saturated or low-signal channels based on an empirically tested range of light intensities. This was confirmed by the administrator before the tests began. The subjects were instructed to avoid large head motions that might induce motion artifacts in the NIRS signal. The NIRS probe remained on the subject as they performed multiple VM trials.

Fig. 1

A 108-channel matrix of NIRS. (a) Shows the four different source–detector separations (15, 30, 36, and 45 mm) with the bottom of the probe aligned with the F7-Fpz-F8 line on the subject’s head. (b) The actual NIRS probe consisted of 15 photodetectors and 12 LED sources.

The NIRS probe was attached to the forehead area because of the lack of hair that could significantly degrade signal quality. The probe consisted of 15 photodetectors and 12 light-emitting diodes (LED) sources. Each of the LED sources alternated between wavelengths of 760 and 830 nm at a temporal frequency of 5 Hz for scanning the entire forehead. The power emitted from the LED sources was $\sim 4\mathrm{mW}$. The 108 channels came from the 108-unique source–detector pairings with different source–detector distances (SD) of 15 mm (40 channels), 30 mm (20 channels), 36 mm (32 channels), and 45 mm (16 channels). The channel matrix is shown in Fig. 1; the green line connecting a source and detector signifies one channel. The separation between source and detector determines the penetration depth, with the greatest sensitivity to gray matter changes being shown at a separation of 30 mm or greater.³⁰ Before the calculation of the NIRS metrics, the data were low-passed filtered (cutoff frequency of 0.2 Hz) and smoothed with a 5-point moving average filter, and the first point of the calculated HbO, Hb, and HbT signals was realigned to zero. In this study, we analyzed changes in HbO, Hb, and HbT to determine whether the observed hemodynamic response varied according to SD.

Resolving for changes in optical density or the log ratio of baseline intensity, $I_o$ and transient intensity $I$ changes at the two wavelengths (760 and 830 nm) allowed accurate measurement of relative changes in HbO, Hb, and HbT,³¹ with HbT being proportional to blood volume.³² In short, the relative concentration change in HbO and Hb can be resolved using the modified Beer–Lambert’s law³³

2.7.

NIRS Parameters

Upon preliminary review of the hemodynamic changes during VM, NIRS parameters were determined to compare the hemodynamic changes between subjects. These parameters were latency, duration, and amplitude calculated for HbO and Hb, and the integrated signal calculated for HbT. Latency was defined as the time between reaching the appropriate pressure to perform the VM ($\sim 40\text{-}\mathrm{mm}$ Hg exhalation pressure) and the beginning of the NIRS hemodynamic response. The amplitude was defined as the maximum concentration change in the calculated NIRS response. Duration was defined as the width of the NIRS response or the time between the start of the NIRS response and when the hemodynamics returned to the baseline state. For HbT, the area under the curve was calculated during the VM exhalation period using the trapz function in MATLAB^®. Because HbT is proportional to blood volume, the integration of this signal can be thought of as sustained blood volume during the VM period. The NIRS parameters were only calculated through the NIRS response obtained from the VM trials suited best for analysis (Fig. 2). The VM trial was selected based on appropriate phase data in the BP measurements. We also used a VM trial that showed an NIRS response with well-defined latency, duration, and amplitude parameters for NIRS analysis.

Fig. 2

NIRS response during VM. Comparison of BP (blue) and HbO (red) changes for (a) an HC and (b) an AbVM patient with OI. In addition, a visual depiction of calculated metrics is shown. (a) The NIRS parameters (latency, amplitude, and duration) and (b) four phases of the VM (according to BP measurements) are shown.

2.8.

Statistical Analysis

Patients were divided into two groups: an NVM group and an AbVM group according to their calculated clinical parameters based on VM results. In addition, data for VM and NIRS in healthy subjects were evaluated. Differences in demographic characteristics, questionnaire scores, and proportion of subjects with abnormal results from VM among the patient groups and the normal healthy group were analyzed using either the Kruskal–Wallis test or Fisher’s exact test. The NIRS parameters were calculated and compared between the two patient groups with OI and normal healthy subjects using a custom-built script written in MATLAB^® (2010A, The MathWorks Inc., Natick, Massachusetts). A significant difference for metrics between the groups was tested using the two-tailed $t$-test. A $p$-value $<0.05$ was considered statistically significant. Pearson’s correlation analyses were performed to identify relationships between VM parameters and the NIRS signal.

3.1.

Subject Characteristics

Data from 25 patients and 3 normal HCs were analyzed. The patients with OI were subgrouped according to VM findings as follows: (1) NVM ($n=9$; median age 66.0; 2 males) and (2) AbVM group ($n=16$; median age 49.0; 9 males). Table 1 presents demographic and clinical characteristics, including VM results, for all subjects.

Table 1

Subject demographic and clinical characteristics.

3.2.

Comparison of NIRS Parameters Between Patients and Healthy Subjects

NIRS parameters (latency, amplitude, duration, and integration) were calculated for the three subject groups: HC, NVM, and AbVM. Figure 2 shows a sample BP and HbO response by (a) HC and (b) AbVM. In this example of an HC and AbVM patient, it can be seen that the AbVM had a longer latency and lower amplitude than the HC; while the HbO peak of the HC occurred during the phase II period of VM, the peak for the AbVM HbO signal came later, and the slope to peak was also lower compared with the HC HbO.

We first investigated whether there was evidence of localized hemodynamic changes across the prefrontal area during VM. The sectioning of the NIRS probe according to its location on the forehead is shown in Fig. 3. The location on the prefrontal area is denoted by the International 10–10 coordinates. Figure 4 shows the box plot for the HbO metrics [(a) latency, (b) amplitude, (c) duration, and (d) HbT integration] for the three groups at a 30-mm SD. It can be seen from the figure that between the six areas, all metrics were measured similarly for the three subject groups. The biggest differences between groups can be seen in the HC group; however, this can be attributed to the low number of subjects. Thus, for simplicity, the proceeding analysis between groups will be shown for all channels averaged across the prefrontal area.

Fig. 3

Probe sectioning used for channel averaging. The NIRS channels were averaged according to six different groups across the forehead. The International 10–10 coordinates are noted on the diagram for orientation.

Fig. 4

NIRS metrics for the six regions: (a) HbO latency for 30-mm SD divided by the six regions (as shown in Fig. 3), (b) HbO amplitude for 30-mm SD divided by the six regions, (c) HbO duration for 30-mm SD divided by the six regions, and (d) HbT integration for 30-mm SD divided by the six regions.

The average HbO and Hb latency for the three groups is shown in Fig. 5. The latencies of the HC and NVM groups were significantly shorter than that of the AbVM group ($p<0.05$). For the HC group, the HbO latency at 30 mm was $0.39\pm 0.23\mathrm{s}$. The NVM group had an HbO latency of $2.79\pm 0.36\mathrm{s}$, and the AbVM group had an HbO latency of $8.14\pm 0.55\mathrm{s}$. For Hb, the latency at 30 mm for HC, NVM, and AbVM was $1.64\pm 0.51\mathrm{s}$, $3.33\pm 0.41\mathrm{s}$, and $5.23\pm 0.34\mathrm{s}$, respectively. Figure 5 shows the statistically significant difference between Hb latency at the 45-mm distance and shorter channels (i.e., 15 and 30 mm). However, the 45-mm distance difference in latency may be attributed to increased scattering of the light reaching that detector. In addition, there were no significant differences between SDs by latency. Overall, our results showed that OI patients had a longer latency for onset of the hemodynamic response than the HC group, with the AbVM group having the longest latency.

Fig. 5

Latency for all averaged channels: (a) HbO latency and (b) Hb latency for all channels are averaged according to source–detector separation. The bars are ordered from left to right from shortest separation (15 mm) to farthest separation (45 mm). Healthy subjects showed the smallest latency at the shorter distance source–detector separations, especially for the HbO signal. Statistically significant values ($p<0.5$) compared with the healthy group or patient NVM group are depicted by * or +, respectively. Statistically significant values ($p<0.5$) between distances within a group are connected via arrows.

While latency clearly distinguished between the groups, amplitude and duration were not significantly different between groups for the NIRS response. The full list of $p$-values between groups for NIRS metrics obtained with a two-tailed $t$-test is presented in Table 2. Although there was no statistical significance, certain patterns can be observed in Figs. 6 and 7. The amplitude of the HbO and Hb response tended to peak at the shorter separations for all groups and decreased as channel distance became longer, as shown in Fig. 6. The drop-off in amplitude is especially noticeable in the Hb response. For amplitude, there was a statistical difference at the 15-mm distance and all other distances for all groups in their HbO and Hb responses. For farther channels, the number of distances that were statistically different was reduced, showing that amplitude was drastically reduced with longer channels.

Table 2

p-values from a two-tailed t-test for groups of patients (healthy, patient normal, and patient abnormal) for each metric (latency, amplitude, duration, and HbT integration) at different hemodynamic measurements (HbO, Hb, and HbT).

Fig. 6

Amplitude for all channels averaged according to source–detector separation. The amplitude of (a) HbO and (b) Hb changes peaked at shorter channels for all patient groups and dropped for the longer channels. Statistically significant values ($p<0.5$) compared with the healthy group or patient NVM group are depicted by * or +, respectively. Statistically significant values ($p<0.5$) between distances within a group are connected via arrows.

Fig. 7

Duration for all channels averaged according to source–detector separation. The bars are ordered from left to right from shortest separation (15 mm) to farthest separation (45 mm) for the recorded NIRS duration of (a) HbO and (b) Hb. The duration parameter shows less difference among the groups. Statistically significant values ($p<0.5$) compared with the healthy group or patient NVM group are depicted by * or +, respectively. Statistically significant values ($p<0.5$) between distances within a group are connected via arrows.

In terms of duration, Fig. 7 shows that HbO and Hb duration did not show any significant pattern between the groups. The difference between channel distances within a group did not demonstrate any recurring trend; however, the 45-mm channels did show the longest duration in the Hb signal for the healthy group.

As a measure of sustained blood volume in the prefrontal area during VM, HbT integration was compared among the three groups (Fig. 8). The HbT integration was significantly different between the HC and AbVM groups. The HC group showed a higher sustained blood volume change in the prefrontal superficial layer during VM compared with the OI groups. Similar to the amplitude metric, the shortest channels were significantly different from other distances for all groups, and the number of statistically significant distances was reduced as the channels became farther apart.

Fig. 8

HbT integration for all channels averaged according to source–detector separation. The bars are ordered from left to right from shortest separation (15 mm) to farthest separation (45 mm) for the integration of the HbT signal. Statistically significant values ($p<0.5$) compared with the healthy group or patient NVM group are depicted by * or +, respectively. Statistically significant values ($p<0.5$) between distances within a group are connected via arrows.

To observe the effect of superficial channels, the shorter channels (15 mm) were regressed from 30, 36, and 45 mm data, using previously described procedures.¹⁵ Figure 9 shows the HbO metrics recalculated after short channel removal. At 30 mm, latency was significantly different ($p<0.05$) between healthy and NVM/AbVM [Fig. 9(a)]. In addition, once the short channels were regressed, HbO duration was significantly different between the healthy and OI patients at 30 mm [Fig. 9(c)]. However, there was no significant difference between NVM and AbVM in the HbO and HbT integration signal when the short channel was regressed.

Fig. 9

Calculated metrics with short channel removed: (a) HbO latency, (b) amplitude, (c) duration, and (d) HbT integration calculated for 30, 36, and 45 mm, with the 15-mm channel regressed from the data. Statistically significant values ($p<0.5$) compared with the healthy group or patient normal group are depicted as a * or +, respectively.

3.3.

Correlation of NIRS Parameters to VM Parameters

We also investigated whether there was any relationship between our derived NIRS parameter and the typical VM parameters used to diagnosis clinical symptoms. We performed a linear regression for the NIRS parameters (HbO, Hb: latency, duration, amplitude, and HbT integration) averaged across the entire prefrontal area, per SD, and the VM parameters (VR, PRT, ${\mathrm{BRSa}}_1$, and BRSv). According to our regression analysis, the only combination of NIRS and VM parameters with a statistically significant relationship was HbO duration at 15-mm source–detector separation and VR ($R^2=0.164$, $p=0.032$). The results of the linear regression are shown in Fig. 10. When HbO duration at 15 mm and VR were tested for each group, there was no significant relationship between the parameters. Results of the linear regression per group were as follows: HC ($R^2=0.003$, $p=0.964$), NVM ($R^2=0.159$, $p=0.288$), and AbVM ($R^2=0.207$, $p=0.076$). The small HC size and similar values in VR between NVM (average: $1.58\pm 0.28$) and AbVM (average: $1.62\pm 0.35$) resulted in a poor fit for each group individually. However, when all the subjects were grouped together, the fit was statistically significant.

Fig. 10

Linear regression analysis of NIRS and VM parameters. Each patient group is designated by a specific shape. This combination of HbO duration (15-mm source–detector separation) and VR was the only combination that showed a statistically significant difference ($p<0.05$).