Shared autonomic phenotype of long COVID and myalgic encephalomyelitis/chronic fatigue syndrome

The study was approved by the Institutional Review Board of the Brigham and Women's Hospital, Harvard University, as a minimal-risk study, and the consent form signature was waived. Authors of the study had access to information that could identify individual participants during data collection.

Orthostatic intolerance was defined as the presence or exacerbation of chronic (>6 months) symptoms attributable to cerebral hypoperfusion – such as lightheadedness, dizziness, dyspnea, brain fog, fatigue and visual disturbances – upon assuming an upright posture, with a partial or complete relief of symptoms upon recumbency. POTS was defined as a combination of orthostatic intolerance and an increment in heart rate ≥ 30 beats per minute for ages > 19 years and 40 beats per minute for ages 18–19 years without orthostatic hypotension during the tilt test [20]. Hypocapnic cerebral hypoperfusion (HYCH), was defined as a combination of orthostatic intolerance and reduced orthostatic cerebral blood flow velocity (CBFv) associated with hypocapnia (end-tidal CO₂< 30 mmHg [21]), but without orthostatic tachycardia or orthostatic hypotension [22]. Orthostatic cerebral hypoperfusion syndrome (OCHOS) was defined by reduced orthostatic CBFv without orthostatic hypotension, orthostatic tachycardia and orthostatic hypocapnia [23].

Patient-reported surveys were done as a part of autonomic testing. The Survey of Autonomic Symptoms (SAS) was used to assess the frequency and severity of autonomic symptoms [32]. The cutoff point > 7 in the SAS score was considered to be clinically significant. Sensory complaints were assessed by the self-reported Neuropathy Total Symptom Score-6 (NTSS-6) [33]. The NTSS-6 total score > 6 was considered as clinically significant. The pain for the last seven days was assessed using the 0–10 numerical rating pain scale (0 = no pain, 10 = worst imaginable pain), a part of the NIH Toolbox [34,35]. The scores ≤ 3 correspond to mild, scores 4–6 to moderate and scores ≥7 to severe pain. Central sensitization was assessed using the central sensitization inventory (CSI), a validated instrument used for evaluation of central sensitization [36]. The score ≥ 40 was used for the diagnosis of Central Sensitization Syndrome.

All testing was performed following established standards and previously described in detail [22]. Medications that may affect autonomic function were discontinued for five half-lives or longer before the testing. Cardiovascular reflex tests included deep breathing, the Valsalva maneuver, and the tilt test. Deep breathing test was performed with inhalation and exhalation each equal to ten seconds which was repeated six times. Parasympathetic cardiovagal index was obtained as the average difference between expiratory and inspiratory heart rate. Valsalva maneuver was performed as a forced expiration at the expiratory pressure 40 mmHg for 15 seconds. The difference between baseline and end of the phase 2 in mean blood pressure was used as a sympathetic adrenergic index. Patients were tilted at 70 degrees for 10 minutes following 10 minutes of supine rest. We described details of autonomic respiratory and cerebral blood flow measurements from the tilt previously [22].

Recorded signals included electrocardiogram, blood pressure, end-tidal CO₂, and CBFv in the middle cerebral artery using Transcranial Doppler. Blood pressure was obtained intermittently every minute by brachial sphygmomanometer using an automated monitor Welch Allyn CVSM 6400 Monitor (Skaneateles Falls, NY) and beat-to-beat using finger cuff the photoplethysmographic signal which was volume-clamped in the finger by servo control (Human NIBP Nano Interface MLA382, ADInstruments Inc., Colorado Springs, CO, USA and Human NIBP Nano Wrist Unit FMS910804, Finapress Medical Systems, Amsterdam, Netherlands). End-tidal CO₂ was obtained using Nonin Respsense Capnograph (Nonin Medical Inc. Plymouth, MN) by nasal cannula. A pulse oximeter (part of Welch Allyn monitor) was used to monitor the oxygen saturation throughout the testing.

The temporal acoustic window with a 2 MHz probe was used to acquire CBFv from the M1 segment of middle cerebral artery using a MultiDop T (Multigon, New York, NY) with an insonation depth between 45 and 65 mm. The transducer has been attached to the head using a head frame with a three-dimensional positioner. The depth and angle of insonation have been kept constant throughout the head-up tilt test. Signals were recorded using the PowerLab 16/35 data acquisition system with LabChart 8 software (ADInstruments Inc., Colorado Springs, CO, USA) and sampled at 400 Hz.

Electrochemical skin conductance (ESC) was used to measure the sudomotor function [37]. ESC correlates with loss of sweat gland nerve fibers and it is a reasonable proxy for sudomotor function [38].

Medical records were also searched for a history of invasive cardiopulmonary exercise testing (iCPET) [39] for evaluation of unexplained fatigue or dyspnea. iCPET was performed in a sitting position using a cycle ergometer as described in detail [40]. iCPET provides Fick cardiac output, right atrial pressure and previously iCPET showed impaired venous return and peripheral oxygen extraction in ME/CFS [18].

Epidermal nerve fiber density (ENFD) and sweat gland nerve fiber density (SGNFD) were obtained at the proximal thigh 20 cm distal to the iliac spine and the calf 10 cm above the lateral malleolus using a 3-mm circular punch tool. Specimen processing including immunoperoxidase staining for the axonal marker PGP 9.5, and fiber counting was done at Therapath (New York, NY) using established standards [41].

SFN is defined as combination of clinical signs suggestive of small fiber dysfunction (pinprick and thermal sensory loss, allodynia, and hyperalgesia) and structural (obtained from skin biopsy) or functional (obtained from QSART or ESC) variables [37,41,42]. The following subtypes of SFN were assessed in this study: sensory SFN (abnormal ENFD, normal SGNFD), mixed SFN (abnormal ENFD and SGNFD), autonomic (normal ENFD, abnormal SGNFD), functional (abnormal electrochemical skin conductance (ESC)) and combined functional-morphological (at least one abnormal: ENFD, SGNFD, and ESC).

Test results were graded using the Quantitative Scale for Grading of Cardiovascular Autonomic Reflex Tests and Small Fibers from Skin Biopsies (QASAT). QASAT is an objective instrument for grading the severity of dysautonomia, small fiber neuropathy, and cerebral blood flow abnormalities that uses normative age and gender adjusted values as appropriate. Each domain (heart rate, blood pressure, cerebral blood flow, end-tidal CO₂) is analyzed separately, where a normal score equals to 0 and the score ≥ 0 is abnormal. QASAT grading is defined as follows [22]:

Autonomic failure score (QASAT_af):

Cardiovagal failure score (QASAT_cardiovagal) is calculated from heart rate responses to deep breathing test. Adrenergic failure score (QASAT_adrenergic) is obtained as a sum of blood pressure responses to the Valsalva maneuver and head-up tilt scores. Sudomotor failure score (QASAT_sudomotor) is obtained from the ESC or QSART. The QASAT_af range is 0–22; none (0), abnormality: mild (1–3), moderate (4–12), and severe (12–22). The additional QASAT ranges are defined as follows: cardiovagal failure: none (0), abnormality: mild (1), moderate (2) and severe (3); adrenergic failure – Valsalva maneuver: none (0), abnormality: mild (1), moderate (2) and severe (3); adrenergic failure – orthostatic hypotension: none (0), abnormality: mild (1), moderate (2–5) and severe (6–10); orthostatic tachycardia: none (0), abnormality: mild (1–2), moderate (3–5) and severe (6–10); sudomotor failure – ESC: none (0), abnormality: mild (1–2), moderate (3–4) and severe (5–6); sudomotor failure – QSART: none (0), abnormality: mild (1–2), moderate (3–6) and severe (7–8); ENFD: normal (0), abnormality: mild (1–2), moderate (3–6) and severe (7–8); SGNFD: normal (0), abnormality: mild (1–2), moderate (3–6) and severe (7–8); reduced orthostatic end-tidal CO₂: normal/none (0), abnormality: mild (1–2), moderate (3–5) and severe (6–10); reduced orthostatic CBFv: normal/none (0), abnormality: mild (1–2), moderate (3–5) and severe (6–10). Details of calculations and grading of the testing were published previously [43].

Both Long COVID and ME/CFS are postinfectious disorders were the autoimmunity and/or low grade inflammation may play a role [4]. Therefore, patient's charts were reviewed for laboratory blood evaluations conducted during routine clinical assessments. We specifically focused on inflammatory, autoimmune, and hormonal markers, as abnormalities in these have been reported in post-infectious disorders [3,44] including: high sensitivity C-reactive protein (normative value <=3 mg/L), tumor necrosis factor-alpha (TNF-α, <= 2.8 pg/mL), interleukin (IL) IL-6 (<7.1 pg/mL), IL-10 (<2 pg/mL), IL-1ß (<7.1 pg/mL), leptin (3.3–18.3 ng/mL), trisulfated heparin disaccharide (TS-HDS) antibody (<10000) titers) [45], fibroblast growth factor receptor 3 (FGFR3) antibody (<3000 titers) [45], acetylcholine receptor binding antibody (<=0.02 nmol/L), ganglionic acetylcholine receptor antibody (<= 0.02 nmol/L) [46], neuronal VGKC antibody (<= 0.02 nmol/L), calcium channel P/Q binding antibody (<= 0.02 nmol/L), myoglobin (<=71 ng/mL) [47] and human growth hormone (0.01–3.61 ng/mL) [47]. We also measured supine (70–750 pg/mL) and standing (200–1700 pg/mL) plasma norepinephrine levels, as these values are useful in assessing the hyperadrenergic form of POTS [20]. The systemic immune-inflammation index defined as neutrophils x platelets/lymphocytes was shown to be useful predictor marker in several malignances [48] was calculated as well.

TS-HDS and FGFR3 antibodies were obtained from Washington University School of Medicine (St. Louis, MO), remaining antibodies as well as TNF-α, IL-10, human growth hormone were obtained from Mayo Clinic laboratories (Rochester, MN). IL-6 was obtained at Mayo Clinic laboratories (Rochester, MN) or at BWH Clinical laboratories (Boston, MA). IL-1ß was obtained at Sunquest (Tucson, AZ). Leptin was obtained at Esoterix Endocrinology (Calabasas Hills, CA). Remaining of laboratory tests were obtained at BWH Clinical laboratories (Boston, MA) or at Quest Diagnostics (Secaucus, NJ).

The continuous data were not normally distributed, and outliers were present. Additionally, the groups failed to meet the assumption of homogeneity of variances. We chose not to remove the outliers, as they may carry clinically important information, and their removal could significantly alter the data distribution. Therefore, we used the nonparametric Kruskal-Wallis test to compare the groups, as it does not assume a normal distribution and is relatively robust to outliers and heterogeneity of variances particularly in large datasets [49]. The effect of size was measured by epsilon squared. If an overall comparison was statistically significant, pairwise post hoc comparisons were performed using Dunn test [50] with the Benjamini-Hochberg adjustment for multiple comparisons [51]. For categorical data, overall comparisons across groups were conducted using the Chi-squared as it does not assume normality [52]. If an overall comparison was statistically significant, pairwise post hoc comparisons were performed using the Fisher's Exact Test with Holm adjustment for multiple comparisons [53].

The effect of head-up tilt on hemodynamic variables was assessed using the linear mixed-effects models adjusted for supine baseline and with a random intercept with repeated-measures design [54]. The predictor variables were the diagnosis and position of the subjects (supine versus upright, minutes 1–10). Gender and age were covariates.

The relationship between lightheadedness during head-up tilt (absent versus present) and QASAT domains was evaluated using the binary logistic regression model.

A higher proportion of missing values was expected in the laboratory blood work, likely due to test ordering being at the discretion of the attending physicians. Ignoring missing data could affect the robustness of our findings. Assuming the data were missing at random, we conducted a sensitivity analysis to assess the potential impact. Specifically, we compared the results of the Kruskal–Wallis test for continuous variables across three scenarios: (1) complete-case analysis using the original dataset with missing data ignored, (2) imputation using the overall mean, and (3) imputation using the overall median.

The R software (www.r-project.org↗) was used for statistical analyses.

The sample size for this study was determined based on a power analysis appropriate for detecting differences among four independent groups using the Kruskal-Wallis test, a nonparametric method suitable for non-normally distributed data. Assuming an alpha level of 0.05 and a desired statistical power of 0.80, the analysis utilized a logistic distribution to approximate the underlying data characteristics. For the primary outcome variable (total QASAT value), based on preliminary estimates of the mean and standard deviation, the minimal sample size required for each group was determined to be 37.2.

Long COVID and ME/CFS had a similar degree of complaints in the autonomic (total and subtotal scores on SAS), neuropathic (NTSS-6), pain, and central sensitization domains (Table 2). hEDS group had worse most of the SAS and NTSS-6 scores, and some central sensitization scores (pain, stiffness) (p = 0.05- < 0.001).

Cardiovascular, cerebrovascular and respiratory variables at supine baseline and during the head-up tilt are shown in the Table 3 and Fig 2A–2D. Comparing all groups in the supine position, there was a significant difference in deep breathing, blood pressure response in the Valsalva maneuver, and ENFD (p < 0.001–0.03) but the difference was not significant in pair-wise comparisons. Furthermore, ESC was higher in hEDS compared to ME/CFS (p = 0.008), and Long COVID (p < 0.001). ENFD at the calf and SGNFD at the proximal thigh were higher in ME/CFS as compared to Long COVID (p = 0.043) and hEDS (p = 0.049). SGNFD was lower in hEDS compared to ME/CFS (p = 0.007). End-tidal CO₂ supine (p = 0.004) and orthostatic (p < 0.001) were lower in Long COVID.

The head-up tilt responses are shown in Fig 2A–2D as minute-by-minute profiles of cardiovascular, cerebrovascular and respiratory variables at rest and during the 10-minute head-up tilt. The patients with Long COVID and ME/CFS had similar heart rate and blood pressure increases during tilt (Fig 2A, 2B), reduced CBFv with a greater than 22% decline from baseline (Fig 2C), and end tidal CO₂ (Fig 2D). There was no difference for all tested variables between Long COVID and ME/CFS, except orthostatic end-tidal CO₂ was lower in the Long COVID group (p < 0.001). The hEDS group had a higher heart rate during tilt (p < 0.03) (Fig 2, Table 3), and higher orthostatic CBFv (p < 0.001) and lower cerebrovascular resistance as compared to Long COVID and ME/CFS. In all subjects, the oxygen saturation was within normal limits throughout the testing (range 96–99%).

Linear model showed significant effect of the diagnoses on the tilt responses for all hemodynamic variables (p < 0.001, Fig 2).

Overall comparisons showed abnormal QASAT scores (>0) in all domains in both groups, indicating mild-to-moderate dysautonomia. Fig 3A–3C and Tables 3 and 4 show absolute QASAT values, % of normalized scores, and frequency of abnormal findings. Both the Long COVID and ME/CFS groups had a worse QASAT_sudomotor score compared to the hEDS group (Figs 3A, 2B).

Orthostatic lightheadedness was observed in >65% of patients, but orthostatic dyspnea was reported only in 21% of Long COVID patients, 37% of ME/CFS patients, and 28% of hEDS patients (Table 4). Tests revealed extensive similarities between Long COVID and ME/CFS, including reduced orthostatic CBFv (80%/92% Long COVID/ME/CFS), mild-to-moderate widespread autonomic failure (89%/95%), presence of SFN (63%/67%), postural tachycardia syndrome (32%/22%) and neurogenic orthostatic hypotension (12%/17%).

Small fiber neuropathy affected ~80–90% of patients using combined morphological and functional criteria. In Long COVID, QASAT_ENFD was abnormal in 48.3% compared to 33.5% in ME/CFS (p = 0.02), but QASAT_SGNFD 27.8% was similar to ME/CFS 28.8%. The rate of SFN (from any biopsy) was similar between Lon COVID 67.2% and hEDS 63.3%,but was higher than in ME/CFS 52.6% (p = 0.04).

iCPET was done in a sitting position, and the results were available in 25 Long COVID and 66 ME/CFS (Table 5). Unadjusted resting stroke volume (Long COVID vs. ME/CFS: p = 0.01), exercise stroke volume (p = 0.01), cardiac output (p = 0.003), and oxygen uptake (p = 0.001) were higher in Long COVID; however, these differences were no longer significant after adjusting for BMI (which was higher in Long COVID). Preload failure was detected in 96% of Long COVID and 92.4% of ME/CFS patients. Deconditioning was present in 64% of Long COVID and ME/CFS patients.

Long COVID and ME/CFS are associated with central sensitization and abnormalities in multiple domains including cerebral blood flow and respiratory dysregulation, small fiber neuropathy, and widespread autonomic failure. Table 6 provides a quantitative summary of main difference between Long COVID and ME/CFS.

Evidence of central sensitization was frequently observed in the majority of our patients. Central sensitization refers to the increase responsiveness of the nervous system to stimuli and is linked to abnormal interoception [55 –57]. The features of central sensitization such as chronic pain, brain fog, fatigue and autonomic complaints has been documented in a variety of pain and fatigue-related syndromes, including [55], long COVID [58], hEDS [59], and chronic fatigue syndrome [56]. Our study confirmed a high prevalence of central sensitization in Long COVID (78.1%), ME/CFS (85.4%), and hEDS (92.4%). The high prevalence of central sensitization in these conditions likely contributes to the significant symptoms burden experienced by patients.

In both Long COVID and ME/CFS, orthostatic CBFv was reduced, due either to abnormal cerebral autoregulation (consistent with OCHOS) or due to hypocapnia-induced cerebral arteriolar vasoconstriction (consistent with POTS and HYCH). Long COVID patients had more frequent orthostatic cerebral blood flow abnormalities and a greater decline in orthostatic cerebral blood flow than ME/CFS patients. Orthostatic hypotension did not play a significant role, because it was only detected in a few patients and orthostatic blood pressure remained in an autoregulatory range.

Reduced CBFv and associated cerebral hypoperfusion may explain some of the disabling symptoms of Long COVID and ME/CFS, such as lightheadedness, brain fog, and chronic fatigue. Previous studies have shown correlations between declines in orthostatic CBFv and lightheadedness [60], a key symptom of cerebral hypoperfusion. Typically a reduction in orthostatic CBFv by 19% or more from the supine baseline is associated with symptoms of central nervous system dysfunction [61,62]. Both our patient groups exceeded that level of decline (Long COVID −25% and ME/CFS −22%). In addition to the reduction of cerebral blood flow, respiratory alkalosis associated with hypocapnia changes neuronal excitability and may alter brain activity [63,64]. Imaging studies using arterial spin labeling showed reduced cerebral blood flow in COVID-19 patients [65,66]. Cerebral hypoperfusion consistent with a large resting state central network dysfunction was detected in Long COVID [67]. Cerebral blood flow dysregulation is also present in hEDS although the abnormality is less severe compared to Long COVID and ME/CFS patients.

Our study detected frequent autonomic failure in both Long COVID (95%) and ME/CFS (89%). Autonomic failure was widespread, affecting cardiovagal, adrenergic, and sudomotor domains. While the cardiovagal and adrenergic abnormalities were mild, sudomotor abnormalities were moderate. Although the dysautonomia tended to be worse in Long COVID compared to ME/CFS, the pattern was similar, affecting multiple domains simultaneously.

Orthostatic intolerance associated with autonomic dysregulation has been observed in both Long COVID and ME/CFS, although the reports are inconsistent [15]. In ME/CFS, most spectral analysis studies showed decreased heart rate variability with reduced parasympathetic and sympathetic activity, with increasing the sympathetic/parasympathetic ratio [13] that was interpreted as relative sympathetic overactivity, although that interpretation is debated [68]. Another study found baroreflex failure with vagal efferent defect derived from analysis of blood pressure variability [15]. Yet another study found no difference in objective autonomic testing compared to chronic fatigue and controls [69]. The variability in findings can be attributed to the heterogeneity of ME/CFS, along with differences in inclusion criteria and techniques for the evaluation. Mild autonomic abnormalities have been detected in Long COVID patients [3,70 –73]. This current study confirmed our previous finding [3,73] and expanded the analysis to the ME/CFS group. Due to the small number of pre-Omicron cases, we did not perform a comparative analysis of the effects of different SARS-CoV-2 strains.

iCPET also showed similarities between Long COVID and ME/CFS including the prevalence of preload failure (96% vs. 92.4%) and deconditioning (64% vs. 64%). Cardiac output which is proportional to BMI, was lower in ME/CFS [74]. However, the Long COVID group had a higher BMI and cardiac output adjusted for BMI was similar between the groups. Although invasive iCPET was available only in a subset of patients, the similarities in key iCPET metrics support the notion of shared common pathophysiology in both conditions.

Skin biopsies, which provide direct evidence of peripheral nerve damage, confirmed the presence of SFN in both Long COVID and ME/CFS patients. SFN was more frequently observed in individuals with Long COVID, although a similar prevalence was also noted in patients with hypermobile Ehlers-Danlos syndrome (hEDS).

We were unable to find particular features in laboratory values that would differentiate the studied disorders. Most of the subjects had normal laboratory values, and abnormal results were found in a minority of patients. There were no differences between Long COVID and ME/CFS in inflammatory, autoimmune, adrenergic, and hormonal markers. These findings are consistent with the hypothesis that both disorders may share a common pathophysiological mechanism. However, the failure to detect elevated inflammatory or autoimmune markers does not support an inflammatory or autoimmune theory for either disorder. Neverthelles, the tests used in our study may not be sensitive enough to detect low-grade inflammation. Elevated cytokines, including IL1β, IL6, and TNFα, have been reported in some, but not all, studies of Long COVID [75]. No differences were found in our study.

Furthermore, we were unable to document differences in levels of norepinephrine, a marker of adrenergic functions, which can be abnormal in peripheral dysautonomia [76]. Nevertheless, norepinephrine levels in peripheral blood do not correlate with central adrenergic activity [77]; therefore, our study cannot rule out central adrenergic dysregulation. Although hormonal dysregulation has been implicated in Long COVID [78], our study did not confirm this finding. We were unable to detect hormonal changes indicative of adrenal or hypothalamic-pituitary-adrenal axis insufficiency, as evidenced by normal cortisol and ACTH levels across our studied groups [79].

While laboratory blood evaluations were unable to distinguish among the three disorders, surveys, and autonomic functional assessments with skin biopsies revealed differentiating features. hEDS subjects reported more severe sensory and autonomic symptoms compared to Long COVID and ME/CFS. Although some overlap was observed, hEDS was associated with less severe cerebrovascular dysregulation but more pronounced peripheral neurodegeneration, as evidenced by greater sudomotor dysfunction and more frequent and severe small fiber neuropathy.

We are a referral center for dysautonomia, so we may not see a representative group of patients. We also used historical controls. However, the large number of patients we have studied may still provide a representative sample of these patient populations.

A limitation of direct Long COVID and ME/CFS comparison is the fact that the duration of the symptoms was much longer in ME/CFS. Duration of the disease may affect the signature of ME/CFS, particularly the immunological profile [80] which was not different between the groups. A greater prevalence of deconditioning, which would be expected with a longer-lasting ME/CFS, was also not detected in our study. Furthermore, autonomic failure was more severe in the Long COVID group, also speaking against the time effect. Nevertheless, it would be useful to longitudinally observe ME/CFS and Long COVID to determine whether these entities converge into an indistinguishable syndrome, which would provide additional evidence about the common pathophysiology of both disorders.

The lack of laboratory values in healthy controls is another study limitation. However, all laboratory tests were validated in a clinical setting, have established normative data and performed at CLIA-certified laboratories.

Methodologically, cerebral blood flow was assessed indirectly using transcranial Doppler, which measures flow velocity, and not flow directly. The velocity is proportional to blood flow, assuming that the diameter of the insonated vessel does not change during orthostatic stress, which was confirmed by an imaging study [81]. CBFv is also affected by the angle of the transcranial Doppler probe. Although the angle varies from patient to patient, once the probe was properly positioned and stabilized with a 3D holder, the same angle was maintained throughout the testing.