An invited Editorial [1] comments on a recent work where we used cardiac-gated phase-contrast magnetic resonance imaging (PC-MRI) to estimate net volumetric cerebrospinal fluid (CSF) flow rate in the cerebral aqueduct of individuals with cerebral aneurysms with or without a previous subarachnoid hemorrhage (SAH) [2]. The authors seemingly find daily CSF flow rates in the order of liters heavy to digest, particularly as they try to interpret the results in light of traditional concepts about CSF production and resorption.

First, we are indeed aware that PC-MRI has methodological weaknesses, which may become even more imminent by net CSF flow measurements. However, our main objective by reporting these data was to make the point that net CSF flow can occur in both directions through the cerebral aqueduct and in very different amounts in different patients. Even though the precision level of our results may be up for discussion, we can hardly see how the critique raised against our methodology changes the essence of the findings by order of magnitude and direction. The objections raised include too large slice thickness (5 mm), use of different scanners, long acquisition time (6 min) making results susceptible to respiration, and application of tailored VENC to measure high CSF flow velocities. We think otherwise; slice thickness of 5 mm (and in-plane resolution of 0.63 × 0.63 mm2) is quite average [3] and should be acceptable at the aqueductal mid-level, where shape is close to being tubular. The fact that our findings are not machine-dependent should be considered a strength, not a weakness. Longer imaging duration renders for averaging out respiration effects, not increased influence. Correctly applied VENC for accurate measurements of high flow velocities in the center of the lumen should be preferred above aliased flow, as high flow contributes more to flow volume than slow flow in the periphery.

However, the main point being made in the Editorial [1] is simply that “bucket flow” does not fit into a model where CSF is produced exclusively by the choroid plexus within the ventricles and resorbed exclusively at the arachnoid villi at the brain surface. Harvey Cushing, by many considered as the father of modern neurosurgery, coined the term “the third circulation” in 1925 [4], in part based on the experiments by his contemporary, Walter E Dandy, who observed hydrocephalus after plugging the foramina of Monro in one single dog [5], and also the experiments by Lewis Weed [6], who had to proceed from small CSF tracers to avoid “diffuse tissue staining” within the brain to arrive at tracers large enough to accumulate solely at brain surface (reviewed by [7]). Since 1925, many lines of research have provided evidence that “the third circulation” represents a profound over-simplification for describing the pathways of CSF. It is, however, understandable that this model, attractive by its simplicity, and in which many scientists have invested much of their work, is hard to abandon.

Today, we know that the subarachnoid space is continuous with the perivascular and interstitial spaces of the entire brain and spinal cord, not only in animals [8,9,10,11], but also in humans [12, 13] (Fig. 1a). To and from the perivascular space, water molecules are continuously exchanged over the capillary wall [7], which in human brain renders for a surface area for water exchange of up to 15–25 m2 [15]. CSF and some of its molecular constituents are excreted to blood over the capillary wall, resorbed by true lymphatic vessels in the wall of dural sinuses [16, 17], and/or drained along lymphatic pathways through neuroforamina at the skull base [18]. In this respect, applying Davson’s equation [19] to estimate to which extent all CSF is resorbed at the arachnoid villi must be considered an anachronism. That said, modeling studies demonstrated that the forces involved in aqueductal CSF flow are orders of magnitude smaller than those predicted by Davson’s equation [20, 21].

Fig. 1
figure 1

The subarachnoid space communicates with the entire extravascular compartment of the human brain (a) and has been confirmed on cohort level in prospective studies [12,13,14]. The color scale illustrates percentage increase in normalized T1 signal intensity in brain of iNPH patient 24 h after intrathecally administered MRI contrast agent (gadobutrol) as CSF tracer. In contrast to normocephalic reference subjects (b), early and persistent CSF tracer reflux is a typical feature of hydrocephalic patients diagnosed with iNPH (c). Compared to pre contrast MRI (d), late scans obtained 24 h later demonstrate persistent ventricular tracer reflux (a) and enhancement of periventricular white matter (a and e)

Full size image

Contrary to reference subjects (Fig. 1b), net retrograde aqueductal CSF flow in patients has been indicated by direct observations of early and persisting ventricular reflux of CSF tracer (intrathecal MRI contrast agent) [12, 13] (Fig. 1c). Ventricular tracer reflux precedes escape of tracer through the ventricular ependyma (Fig. 1a, d, e), even though the molecular size of tracer is far above that of water. In this sense, assuming that ventricular CSF would in its entirety be drained by an inserted tube appears as a logical shortcut.

A major issue of general interest in science is observations being put aside as flawed when not in line with a predefined model. Thereby, bias is introduced as the established model is exclusively receiving support. Science should rather be data-driven, not governed by hypotheses that are almost predefined as unalterable. According to the science philosopher Karl Popper (1902–1994), true progress in scientific knowledge goes through the method of falsification rather than verification, or “enlarging the graveyard of falsified hypotheses” (reviewed by [22]). This would, however, require both original and independent thinking.