Cerebral blood flow is reduced in ME/CFS during head-up tilt testing even in the absence of hypotension or tachycardia: A quantitative, controlled study using Doppler echography

Executive summary

This study measured blood flow to the brain in 429 people with ME/CFS and 44 healthy people while they were tilted upright on a table for up to 30 minutes. It used ultrasound of the four arteries in the neck that supply the brain, measuring both how fast blood moved and how wide the vessels were. That gives the volume of blood reaching the brain, which is better than older methods that measured speed alone.

Healthy people lost about 7% of their brain blood flow when upright. People with ME/CFS lost about 24–29%. This was true even in patients whose heart rate and blood pressure stayed normal on the tilt test. The difference is large, the flow calculations were done by someone who did not know who was a patient, and lying-down flow was the same in both groups, so the main finding is probably real.

But the study has clear limits. The healthy controls were not matched for fitness, so it cannot say whether the larger fall is specific to ME/CFS or would be seen in anyone very inactive or chronically ill. The person who recorded every scan knew who was a patient. Patients breathed off more carbon dioxide when upright, which by itself narrows brain blood vessels, and the analysis did not account for this. About 30% of patients met only a case definition that does not require post-exertional malaise. The link between blood flow and symptoms was moderate, not "strong" as the paper says, and does not show that low flow causes the symptoms. The cut-off used to call 90% of patients "abnormal" was set on only 44 healthy people and has not been tested against other illnesses.

The central finding holds; the claims about cause and diagnosis built on it are weaker than stated.

Grade: 6 / 10.

Study design

This is a single-centre study of patients seen at the Stichting CardioZorg, a cardiology clinic in Hoofddorp, the Netherlands, that specialises in ME/CFS. Between October 2012 and January 2018, 714 consecutive patients came to the clinic with suspected ME/CFS. Those diagnosed with ME/CFS were usually given an orthostatic stress test as part of their clinical work-up. The paper analyses the 429 who had a head-up tilt test (HUT) and usable measurements, and compares them with 44 healthy controls.

The test: 20 minutes lying flat, then tilted head-up to 70 degrees for up to 30 minutes. Heart rate and blood pressure were recorded continuously with a finger cuff (Nexfin). Blood flow was measured with ultrasound in the four arteries that supply the brain: both internal carotid arteries and both vertebral arteries. For each artery, flow was calculated as mean flow velocity times the cross-sectional area of the vessel, and the four were added to give total cerebral blood flow (CBF) in ml/min. Measurements were taken lying down, at mid-tilt (starting at about 12 minutes) and at end-tilt (starting at about 22 minutes). End-tidal carbon dioxide (PetCO2, the carbon dioxide level in breath at the end of breathing out) was monitored.

This matters because most earlier work used transcranial Doppler, which measures flow velocity in arteries inside the skull. Velocity only tracks flow if the vessel diameter stays the same, and diameter can change, for example when carbon dioxide falls. The authors measure both velocity and diameter in the neck, so they measure volume flow, not just velocity. The guide asks this question for cerebral blood flow studies, and this paper answers it well.

Patients were grouped by their heart rate and blood pressure response to tilt, using the consensus criteria: normal response (247), delayed orthostatic hypotension (dOH, a fall in blood pressure that comes after the first 3 minutes upright; 62), and postural orthostatic tachycardia syndrome (POTS, 120).

From August 2013, a clinician read out 15 yes/no questions about symptoms at the first minute of tilt and again at the 10th minute.

There was no pre-registration and no sample size calculation. The Discussion says "we prospectively studied OI [orthostatic intolerance] in 429 adults", but in practice these are tests done in routine clinical care and analysed afterwards. The paper says some tests were chosen for their use in social security claims. The 15-question symptom questionnaire was added part way through, after a pilot. The comparisons themselves are simple and few, so the room for choosing analyses after seeing the data is limited.

Sample selection

Case definition and post-exertional malaise

One clinician (Visser) made every diagnosis at the first visit. All 429 patients met the Fukuda criteria for chronic fatigue syndrome, and 301 (70%) also met the 2011 International Consensus Criteria for ME of Carruthers and colleagues. The Fukuda criteria do not require post-exertional malaise (PEM). The International Consensus Criteria do. So up to 30% of the sample may not have PEM as a required feature. The paper does not say how PEM was assessed, and no questionnaire such as the DePaul Symptom Questionnaire was used. The paper says exclusion criteria were "taken into account", but does not describe the work-up: which blood tests, whether sleep disorders or psychiatric conditions were screened for, and how.

The results are not reported separately for patients who met the International Consensus Criteria and those who met Fukuda only. The guide suggests that a signal specific to ME/CFS should get stronger under stricter criteria. That check was possible with these data and was not done.

On the other hand, this is a specialist clinic with one experienced diagnostician, and only 39 of 714 referred patients were judged not to have ME/CFS. The share of true ME/CFS cases is probably higher than in a self-report sample, but it cannot be estimated from what is reported.

Orthostatic intolerance, POTS and mast cell activation

Orthostatic intolerance is the subject of the study, so it is well described. Daily-life orthostatic symptoms were judged by the clinician at interview, and a second clinician reviewed 60 randomly chosen charts (intraclass correlation coefficient 0.80). 86% of patients reported daily-life orthostatic symptoms, against 3 of 44 controls (7%). POTS was diagnosed on the tilt test itself. Mast cell activation syndrome (MCAS) is not mentioned.

Severity and who was left out

Of 675 patients diagnosed with ME/CFS, 165 did not have a tilt test (Figure 1). 100 had an alternative standing or sitting test because they were judged more severely affected, 31 had an exercise test instead, 9 were too impaired to be studied, 15 refused, and 10 were evaluated elsewhere. Bedbound patients were not enrolled. So the most severely ill are under-represented, which the authors acknowledge. Severity was not measured with a scale. Median disease duration was 10 years (interquartile range 5–16).

After the tilt test, 39 patients were excluded because the Doppler images were of insufficient quality, and 27 because they took drugs that lower heart rate or blood pressure. A further 15 were not analysed because blood pressure fell too fast (7 with classic orthostatic hypotension) or near-fainting came on too quickly (8) for imaging. Those 15 are likely to be people with large falls in cerebral blood flow, so leaving them out works against the authors' finding, not for it. Figure 1 shows no control excluded for image quality: 3 of 47 were not analysed, one with delayed orthostatic hypotension and two with rapid near-fainting.

Seven patients suspected of POTS stopped beta-blockers or ivabradine for 7 days before the test. Other medication was continued. Patients on heart rate or blood pressure lowering drugs were excluded.

The authors also acknowledge referral bias: general practitioners may have sent patients with orthostatic symptoms in particular to a cardiology clinic. This affects the prevalence figures (the "90%") more than the comparison with controls.

Controls

The 44 healthy controls were recruited through ME/CFS patient advocacy websites, posters in the clinic's building, and healthy acquaintances of patients. They match the patients on age, sex, body mass index and body surface area (Table 1). Their activity level and fitness are not reported. They are described as healthy, not as sedentary or deconditioned.

This is the main gap in the comparison. Patients with ME/CFS are on average far less active than healthy people, and lack of activity on its own reduces blood volume and orthostatic tolerance. Bed rest is a standard way to induce orthostatic intolerance in healthy volunteers. Without a deconditioned or disease control group, the study cannot say whether the larger fall in cerebral blood flow is specific to ME/CFS or would be seen in anyone equally inactive or ill. The paper does not discuss deconditioning.

Lying-down cerebral blood flow did not differ between patients and controls (615 versus 625 ml/min), which argues against a simple difference in measurement between the groups.

The controls were studied at an unknown time relative to the patients. A control group of 44 against 429 patients is also small for setting a "lower limit of normal" (see below).

Outcome measures

Cerebral blood flow

The guide counts cerebral blood flow as an objective measure, and volume flow measured in the neck arteries is a sound choice. A patient cannot easily change it by effort. But it is not free of observer influence.

• Acquisition was not blinded. One operator (Visser) recorded all the images. He was also the treating clinician, made the diagnoses, and read out the symptom questions during the same test. He knew who was a patient and who was a control.
• Analysis was blinded. The flow calculations were done offline by a second investigator (van Campen) who was "unaware of the patient or control status and unaware of the hemodynamic outcome". That is a real strength. The paper does not say how the images were anonymised. The heart rate is visible in the Doppler trace, so an analyst could in principle recognise the faster rate of POTS, though not the difference between patients with a normal response and controls, whose heart rates at end-tilt were similar (88 and 83 beats per minute).
• Diameter errors count twice. Flow is velocity times area, and area goes with the square of the diameter. A 5% error in diameter gives about a 10% error in flow. The operator chooses where and how to image the vessel, so the unblinded acquisition matters more for this method than it would for a blood test.
• Reliability was shown only in healthy people. The paper cites the group's 2018 study in healthy volunteers for good intra- and inter-observer variability. Reproducibility in patients, and day-to-day variation, are not reported. The authors themselves say the method "has to be replicated by others".

The symptom questionnaire

The 15 questions were read out by the same clinician who was doing the imaging, and answered yes or no. The questionnaire was developed by the group and has not been validated. Some items, such as fatigue, leg weakness and concentration, are common in ME/CFS whether or not the person is upright, though the questions do ask about change since tilting. Answers were summed into a score from 0 to 15. Because the person asking the questions knew the patient's group and could see the test, the symptom scores are open to both patient and observer expectation.

End-tidal carbon dioxide

PetCO2 was measured, but the paper reports it only as group medians and uses it once, to split patients at 30 mmHg (Figure 3). It is not used to adjust the blood flow comparison. This matters a lot; see the next sections.

Bias

The authors run a cardiology clinic that specialises in ME/CFS, and they developed this measurement method themselves (their 2018 study in healthy volunteers). They have an interest in the method working. That is the setting in which the measuring clinician knew each participant's group. The analysis was blinded, which limits the damage, but the acquisition and the symptom questions were not. The paper does not say who judged which images were good enough to use, or whether that judgement was blinded.

The authors declare no conflicts of interest. Rowe is supported by the Sunshine Natural Wellbeing Foundation Professorship. No other funding is stated.

Statistical and methodological issues

The size of the difference

The group difference is large. At end-tilt, controls lost 6.7% (standard deviation 3.1%) of their lying-down cerebral blood flow. Patients with a normal heart rate and blood pressure response lost 23.6% (10.3%), those with delayed orthostatic hypotension 27.7% (6.0%), and those with POTS 28.9% (5.8%) (Table 2). Across all patients, flow fell from 615 to 456 ml/min at end-tilt. Figure 3 shows the 95% confidence intervals for controls and patients well apart. A difference of this size is not plausibly a product of multiple testing.

Carbon dioxide is a large confounder

At end-tilt, median PetCO2 was 36 mmHg in controls, 32 mmHg in patients with a normal response, 32 in delayed orthostatic hypotension and 26 in POTS (Table 2). Breathing out more carbon dioxide than normal (hypocapnia) narrows brain blood vessels and reduces cerebral blood flow. The authors know this: it is their own argument against transcranial Doppler, and Figure 3 shows a larger fall in patients with end-tilt PetCO2 of 30 mmHg or lower (about 29%) than above 30 (about 24%).

So part of the difference between controls and patients with a "normal" response could come from patients breathing more while upright, rather than from a failure of blood supply. The paper does not adjust for PetCO2, does not report the relation between PetCO2 change and flow change within groups, and in its limitations section says a dose-response with end-tidal carbon dioxide would strengthen the findings. Even patients with PetCO2 above 30 mmHg had a mean fall of about 24%, much more than controls, so carbon dioxide does not look like the whole explanation. But how much of the gap it explains cannot be told from what is reported.

This does not make the finding false. Hypocapnia during standing would itself be a real abnormality; the authors cite Novak (2018), who described "hypocapnic cerebral hypoperfusion" as a marker of orthostatic intolerance. It does mean the title's contrast, reduced flow "even in the absence of hypotension or tachycardia", should not be read as reduced flow without any measurable cause.

The "lower limit of normal" and the 90% figure

The authors set a lower limit of normal at the control mean plus 2 standard deviations of the end-tilt fall, which gives 13%. By that cut-off, 384 of 429 patients (90%) had an abnormal fall: 82% with a normal response, 98% with delayed orthostatic hypotension and 100% with POTS.

Three problems:

1. A cut-off set from 44 people is imprecise. The control spread (standard deviation 3.1%) is narrow, and a slightly wider spread in a new control sample would move the cut-off.
2. The cut-off was set and tested on the same controls, so its specificity has not been tested in an independent group.
3. There were no disease controls, so "abnormal" means "different from these 44 healthy people", not "specific to ME/CFS". The 90% is a prevalence in a referred clinic population that excluded the most severe, not a diagnostic accuracy.

The 90% figure is a fair summary of how far patients differ from these controls. It is not evidence that the test can diagnose ME/CFS or orthostatic intolerance, and the paper's "Significance" paragraph says it "may have implications for the diagnosis" without that caveat.

The symptom correlation

Figure 4 plots the number of symptoms at 10 minutes against the fall in flow at mid-tilt. The figure shows R = 0.40 and the line Y = −0.13X + 3.54. That is about one extra symptom for every 8 percentage points of extra fall in flow, and the fall in flow accounts for about 16% of the variation in symptom count (R squared). The paper calls this "the strong association" in the Discussion and says the reduction is "strongly associated" with symptoms in the Conclusion. A correlation of 0.40 is moderate.

The paper does not say how many people are in Figure 4 or who they are. The questionnaire started in August 2013, so it cannot include all 429. The text calls it "per patient", but the cluster of points near 0% fall with 0 to 4 symptoms looks like the controls, who had a mean fall of 5% and a mean of 0.5 symptoms. If controls are included, much of the correlation would come from the difference between groups rather than from variation among patients. This cannot be checked from the paper.

Symptoms and flow were also recorded by the same unblinded clinician during the same test.

Biological plausibility

A fall in cerebral blood flow on standing is expected in people with POTS or delayed orthostatic hypotension. The new claim is that patients with a normal heart rate and blood pressure response lose nearly as much. Two mechanisms could produce that without either measurement being abnormal: a fall in cardiac output (less blood pumped per minute) that is compensated in blood pressure by tightening of blood vessels, and hypocapnia. The authors name both. Neither was measured here in a way that could separate them; cardiac output was not measured at all.

The claim fits with a strong practical observation: many people with ME/CFS feel worse upright without meeting POTS or hypotension criteria. It would predict that cardiac output falls more on tilt in patients than in controls, that correcting carbon dioxide (for example by adding it to inhaled air) would partly restore flow, and that treatments that increase blood volume would reduce the fall. The paper itself tests none of them.

Interpretation and spin

Most of the paper's claims match its data: patients' cerebral blood flow fell more during tilt than controls', including in the group with a normal heart rate and blood pressure response. The abstract's conclusion is restrained: "extracranial Doppler measurements demonstrate that CBF is reduced in ME/CFS patients with POTS, dOH, and even in those without HR/BP [heart rate and blood pressure] abnormalities."

Some statements go further than the data:

• "It clearly demonstrates that reduced cerebral blood flow is a cardinal contributor to orthostatic intolerance symptoms." The paper draws this from the larger fall in flow in patients with delayed orthostatic hypotension and POTS than in controls. A difference between groups, like the moderate correlation with symptoms, cannot show that reduced flow causes the symptoms. It is consistent with that, and with both being driven by something else, such as hyperventilation.
• "The clinical importance of this observation is reflected in the strong association" and "strongly associated". The correlation is 0.40.
• "These patients would have been misclassified as having normal hemodynamics", and "cerebral blood flow measurements should be included" in categorising circulatory dysfunction. This assumes the 13% cut-off marks disease, which has not been validated against disease controls or outcomes.
• "We prospectively studied OI [orthostatic intolerance] in 429 adults". These were consecutive clinical tests analysed together; there was no prospective protocol.
• The Highlights state that cerebral blood flow reduction "correlates with symptoms of orthostatic intolerance", which is fair.

The limitations section is honest about referral bias, the exclusion of the most severe, the lack of a dose-response with carbon dioxide, and the need for replication by others. It does not mention the unblinded acquisition, the lack of deconditioned or disease controls, or the unvalidated questionnaire.

What this demonstrates

In a large clinical sample of people diagnosed with ME/CFS at one specialist clinic, measured by one operator, cerebral blood flow fell about 24–29% during 70-degree head-up tilt, against about 7% in 44 healthy controls. The difference held in patients whose heart rate and blood pressure responses were normal. Lying-down flow was the same in both groups. The flow calculations were blinded. The size of the difference makes chance an unlikely explanation.

It does not show:

• that the larger fall is specific to ME/CFS, rather than to inactivity, deconditioning or chronic illness;
• how much of the fall comes from hypocapnia;
• that the fall causes the symptoms;
• that a 13% cut-off is a valid diagnostic threshold;
• that other operators or centres get the same result;
• anything about the most severely ill, who were mostly not tilted.

Critical missing elements

• A deconditioned or sedentary control group, and a disease control group with another chronic illness.
• Analysis of blood flow adjusted for end-tidal carbon dioxide, or the within-person relation between the two.
• Results split by case definition (International Consensus Criteria versus Fukuda only).
• How PEM was assessed, and the patient's PEM state on the day of testing.
• The number and identity of participants in the symptom correlation (Figure 4).
• Reproducibility of the method in patients, including day-to-day variation.
• Blinded image acquisition, or at least a description of how images were anonymised for analysis.
• Cardiac output or stroke volume during tilt.

Context beyond the paper

A single web search found further tilt studies using this method from the same group, including a larger study of the relation between cardiac output and cerebral blood flow in 534 patients, and studies at a lower tilt angle in severe patients. It did not find an independent replication by another group using extracranial Doppler during tilt. That was one search, not a systematic check, and I did not open those papers.

Recommendations

For readers and patients. This is good evidence that cerebral blood flow falls more during tilt in many people with ME/CFS than in healthy controls, including in people whose standard tilt test looks normal. A normal heart rate and blood pressure response on tilt does not rule out a circulatory problem when upright. It is not yet evidence that this test diagnoses ME/CFS, or that the fall in blood flow explains the symptoms.

For clinicians. The 13% cut-off comes from 44 healthy people at one clinic and has not been tested against other illnesses or in other centres. Treat it as a research measure, not a diagnostic threshold.

For researchers. The most useful next study would repeat this with blinded acquisition, at a second centre, with sedentary and disease controls, and with blood flow analysed against carbon dioxide and cardiac output in the same people. Reporting results by case definition would show whether the signal tracks ME/CFS or orthostatic intolerance in general.

Final assessment

The core finding is probably real. The measure is better than transcranial Doppler because it captures volume flow, the analysis was blinded, the sample is large, lying-down flow did not differ between groups, and the gap between patients and controls at end-tilt is very large. Excluding the patients who could not be imaged because they fell too fast or nearly fainted works against the authors.

What prevents a higher score: the only controls are healthy, of unknown fitness, and few, so the study cannot tell an ME/CFS-specific problem from the effects of inactivity or illness. The operator who recorded every image knew each participant's group, in a method where small errors in vessel diameter double in the flow figure. The large group difference in end-tidal carbon dioxide is left unadjusted, so the title's contrast "even in the absence of hypotension or tachycardia" may be partly explained by hyperventilation. The case definition did not require PEM for 30% of patients. The paper calls a moderate correlation with symptoms "strong", claims a group difference "clearly demonstrates" that low flow is "a cardinal contributor" to symptoms, and it presents a cut-off set on 44 controls as grounds for changing how patients are classified.

What prevents a lower score: none of these flaws could plausibly produce a difference of 17 percentage points between patients with a normal response and controls on their own, and the paper's main abstract conclusion is stated carefully. The central claim, that cerebral blood flow falls more on tilt in this population than in healthy people, holds. The claims built on it — about cause, specificity and diagnosis — are weaker than stated.