Cerebrospinal fluid dynamics in pediatric pseudotumor cerebri syndrome

The pseudotumor cerebri syndrome (PTCS) has been principally described as a disease of post-pubertal females with raised BMI and its accurate diagnosis is particularly challenging in the pediatric population. The current intracranial pressure (ICP) thresholds in the pediatric spectrum of the disease (28 cm CSF for obese and/or sedated children, 25 cm CSF for normal weight and non-sedated children for definite PTCS) [1‐3] have been established epidemiologically, being set at the 90th percentile level of ICP recordings obtained in a “normal” pediatric population [4]. There are no adequately powered, randomized studies to reliably confirm that the thresholds are clinically relevant [1, 5‐7].

There is a growing body of evidence highlighting the importance of long-term ICP monitoring against single “snapshot” manometric measurements. For example, it is well-known that standard lumbar puncture (LP) measurements can be affected by posture, movement, emotion, anesthetic agents, and hypercapnia, which can lead to unreliable estimation of ICP (CSF pressure (CSFp) if it is derived from connecting to the spinal canal rather than the parenchyma) [8‐11]. It is particularly important to bear in mind that ICP levels are dynamic [7, 9‐11] so a “normal” recorded ICP at a single timepoint may not be a representative value. In adults and children, the assessment of average ICP over more than 20 min (“steady state”) is reported to be more reliable than a single opening pressure measurement [5, 10, 12‐14] (Fig. 1). Although it is important to estimate ICP more reliably, baseline ICP as a solitary number cannot provide a full assessment for the diagnosis and understanding of the CSF circulatory disorders. This is because it is the production, circulation or reabsorption of CSF that makes ICP dynamic and is likely to be deranged in PTCS. Historically, only around 30% of these disturbances have been shown to be reflected on steady-state ICP values [7, 15‐20]. For these reasons, monitoring of ICP in conjunction with CSF dynamics has long been established in the field [9, 11, 21‐23].

At our center, children with suspected PTCS routinely undergo a lumbar infusion study. During the infusion test, they have a LP with measurement of a baseline, longer term CSFp monitoring, and, if the pressure is below approximately 30 mmHg, sterile artificial CSF fluid is infused via the spinal needle. The resultant pressure-volume data can be analyzed to give the CSF dynamics variables such as the pulse amplitude of ICP, the compensatory reserve, assessed mainly as the moving correlation coefficient between mean CSF pressure and pulse amplitude (RAP), and the magnitude of slow waves of CSFp [24‐26].

Using this technique, children with optic disc edema at our center can be diagnosed accurately with a raised CSFp (definite PTCS) and normal CSFp (probable PTCS more reliably than with standard LP), and we can also characterize the CSF dynamics of each child.

The methodology described has been used in our center and others to report on the profiles of CSF dynamics that are seen in various types of pediatric and adult hydrocephalus [5, 27‐30]. By contrast, there is little reported information on the profile of CSF dynamics that characterizes PTCS. In adults with PTCS, it has been reported that there is a dynamic coupling between CSFp and sagittal sinus pressure (SSp), which was demonstrated by performing infusion studies simultaneously with direct cerebral venous pressure monitoring via a catheter [31]. Our previous paper [32], which reported on children referred to our center from 2006 to 2009, remains as the only report on CSF dynamics in pediatric PTCS.

We aimed to investigate the profile of CSF dynamics in pediatric PTCS and to see if it could provide additional, clinically useful insights. For this purpose, we performed a retrospective review of the CSF dynamics for all patients referred to our tertiary pediatric services (≤ 16 years old) for consideration of a diagnosis of PTCS from 2006 to 2016. We have re-analyzed their original CSF infusion study data in an up-to-date mathematical model and investigated the CSF dynamics disturbances in definite and probable pediatric PTCS. We have also tested their diagnostic yield compared with the traditional investigations that these children undergo.

The diagnostic implications of measuring the full set of CSF dynamics parameters in pediatric PTCS are highlighted in this paper, in order to provoke further investigation and to encourage the use of this approach more widely in clinical practice. CSF dynamics could provide information on the contents of CSFp that are not obvious from CSFp as a solitary number, accurately differentiating between the clinically and radiologically derived entities of PTCS.

We have highlighted that (A) a single recording of a raised CSFp is not necessarily a representative value for a patient’s ICP and (B) a raised CSFp value does not give a full understanding of a patient’s CSF circulation, including venous drainage. Despite its limitations, we acknowledge that CSFp is a diagnostic criterion in current guidelines and we have therefore described our precise methodology for maximizing the reliability of this single CSF dynamics variable. This methodology includes the measurement of “steady-state” CSFp ([8, 9, 42]) and simultaneous measurement of the blood circulation with CSF pressure recordings, as it is known that there is a complicated interaction between the CSF circulation and the cerebral blood circulation in CSF disorders ([7, 9, 43]). We also aim to minimize the impact of other confounding factors such as stress, position, and sedation [1, 4, 11, 47]. These factors and how they influence the CSFp cannot be studied and comprehended without analysis of the CSFp components, preferably with cerebral multi-modality monitoring [11, 25, 42].

Studies have already shown that there is a lack of reliability in the manometry reading of a CSFp [8, 11], something which is also partially shown by the lack of correlation between manometry and baseline CSFp from a computerized infusion test in our cohort. However, these readings were not taken at a very close time, but up to 3 months apart. Nonetheless, they represent the clinical reality of trying to make a diagnosis and a treatment plan from separated LPs and/or drainages, often weeks and month apart from each other.

The CSFp at baseline was by definition raised > 20 mmHg in pediatric PTCS and on average in our group > 25 mmHg. In the group of children with definite PTCS, there were 3 children with a CSFp > 30 mmHg, and 2 of them had critical levels of CSFp with compromised cerebral perfusion. Although infusion was not possible in these children, the CSF dynamics parameters that could be measured were revealing; a pattern of negative AMP-CSFp relationship was observed, indicating low-cerebral perfusion pressure. It has been reported in traumatic brain injury that when ICP reaches critical levels (usually > 25 mmHg in head injury), the AMP-ICP relationship becomes negative [9, 11], which appears to also be the case with these 2 patients. However, more data are needed in order to confirm this pattern and association in pediatric PTCS. These cases, although seemingly rare, highlight the importance of an extended multi-parametric monitoring of the cerebral circulation, including systemic arterial pressure and possibly cerebral autoregulation, blood flow, and oxygenation. It would be interesting, for future purposes, to adopt at a research setting initially a multi-parameter monitoring approach, especially in suspicion of serious disease and explore the clinical implications of possible cerebral hypoperfusion.

The difficulty in diagnosing and classifying pediatric PTCS lies in the probable group, where the CSFp was not raised (average 15 mmHg) and did not differ to the children in whom PTCS was excluded. In these cases, it is of significance to investigate the dynamics of the CSFp. Neither CSFp nor AMP at baseline seemed to help in differentiating between those probable and not PTCS. However, both elasticity and SSp were significantly lower in group C and could potentially provide a key in the differentiation between the two groups. Both elasticity and SSp were elevated in children with definite as well as probable PTCS, showing further potential markers of the disease besides CSFp. However, it is important to bear in mind that both of these parameters are subject to calculation errors and SSp represents a value derived from mathematical modeling of the CSF circulation and not the patient’s actual SSp. Unfortunately, it is not possible to understand or confirm what the probable PTCS group represents, as compared with definite PTCS and normal children, until at least some further follow-up and perhaps investigations are performed. Higher SSp as well as higher elasticity could perhaps highlight a tendency or the beginning of the syndrome that has not reached the severity of the definite group.

Besides elasticity and SSp, and partially RAP, none of the other infusion-derived parameters were above the reported normal thresholds (0.6 for RAP—lowered to 0.4 for the this cohort of patients, 0.16 for the amp-p line, 4 mmHg for AMPp, 10–13 mmHg*min/ml for Rout), with the exception of slow waves of CSFp at plateau, the magnitude of which exceeded 1.5 mmHg in groups A and B, but not in group C. Also, the absolute rise in CSFp during infusion was the same in all groups. This rise, divided by infusion rate, denotes the resistance to CSF outflow. As PTCS SSp follows the rise in CSFp, this way of estimation gives an overestimated value of resistance. Nevertheless, in PTCS, CSF circulation is probably normal, as the gradient CSFpp–CSFpb is low [15].

For slow waves in particular, there is no well-described threshold in any of the CSF disorders [24, 26, 48]. Although the differences in the magnitude of slow waves between groups were not significant, there seems to be a tendency of higher magnitude in group A than in groups B and C, group C possibly having the lowest magnitude. Perhaps an increased magnitude of slow waves when the CSF circulation is challenged could reveal a disturbed CSF circulation and assist in differentiating between definite, probable PTCS and mimics. This will need further confirmation. Another finding that could provide interesting information in excluding or confirming PTCS is the presence or absence of a LBP in the AMP-P line. None of the children in the definite group presented with such a breakpoint, whereas it was observed in 2 children for each of the other groups.

There was a correlation between CSFpb and SSp, but not CSFpp. This finding could be similar to what has been shown in adults with PTCS [31], whereby SSp and CSFp are coupled at baseline and during infusion. This relationship, however, cannot be validated in children without direct SSp measurement and there is ethical equipoise currently between the risks and benefits of exposing them to invasive cerebral venous investigation. The correlation between SSp and elasticity found mainly in our definite PTCS patients could also demonstrate part of the pathophysiology of PTCS, whereby “faulty” venous sinuses allow the transmission of the CSF pressure onto the sinuses, leading to a coupling of the two pressure and thereby a “stiffer” brain. It is therefore possible in children with normal CSFp but abnormal CSF dynamics (e.g., a depleted compensatory reserve) or an abnormal cerebral venous sinus system [5] that CSFp levels can rise with minimal perturbation of the current equilibrium state.

It was intriguing to find out that no single CSF pressure or dynamics parameter on its own could accurately differentiate between the 3 groups with an AUC > 0.80. Our findings from the ROC analysis show a potential overlap between diagnosis and classification from clinical examination and neuroradiology, with CSF dynamics parameters, in particular, CSFpb, SSP, and elasticity. The main discrepancy between CSF dynamics and the Friedman classification lie in the patient in group C with no papilledema and disturbed CSF dynamics and in group B, where there is a lot of heterogeneity among the patients. We therefore could speculate that monitoring and infusion, not just baseline parameters, are possibly essential for accurate patient classification and characterization of the disturbance in the CSF circulation. In this context, it is of interest to look into the case of one out of the 5 patients without PTCS, whose CSFp was 27 mmHg and did not meet the criteria for PTCS from the Friedman classification. Therefore, in the absence of papilledema, even a confirmed intracranial hypertension perhaps is not synonymous to the actual syndrome and the venous compartment in such cases could merit further investigation. Moreover, it may be possible to distinguish between these two entities both clinico-radiologically, as well as with monitoring of CSF dynamics, and more patients will be needed to make these distinctions.

Finally, the influence of GA on CSF dynamics could not be clearly demonstrated in this small cohort. It has been reported in NPH and TBI patients that GA possibly has no effect on baseline CSFp or elasticity, but significantly dampens the magnitude of slow waves [25]. This study was derived from the same center and therefore included the same GA protocol as the NPH group (propofol + remifentanil and a muscle relaxant, usually rocuronium, naturally with different doses in adults vs pediatric patients). The fact that CSFpb was increased in definite PTCS under GA vs conscious children could be a random finding, or show that anesthetic agents could increase CSFp in pediatric PTCS patients, which is not justified from the literature on the influence of anesthetic agents such as propofol on cerebral blood flow and metabolism. Additionally, this finding, as well as similar previous reports [4, 25, 47], could highlight that children needing GA are usually the ones with a higher BMI, that in our experience is less likely to tolerate a LP, and could reflect the known relationship with BMI in PTCS severity as well as the influence of increased abdominal pressure on CSFp. Preliminarily derived from our set, 7/14 children needed GA in group A, 3/3 with CSFp> 30 mmHg. In comparison, our adult and pediatric hydrocephalus children normally require GA when their symptoms are very prominent, which prohibits them from complying with investigations. Similarly, the lower SSp and elasticity in anesthetized probable PTCS patients could mean several things, but this study is not adequately powered to identify if this effect is real. Notably, 14/39 (36%) of our total children (including the ones diagnosed with secondary PTCS and were excluded from this analysis) that underwent an infusion test had GA during the investigation. This percentage was lower compared with a reported national average of 45% [2], and we utilized our longer term average CSFp technique that allowed for reliable CSF pressure measurement on a truly continuous scale.

CSF dynamics in PTCS show a peculiar profile: baseline CSF pressure is elevated and estimated sagittal sinus pressure is elevated. Elasticity and compensatory reserve are depleted. CSF circulation is probably undisturbed. Very high CSF pressure may contribute to low cerebral perfusion, exposing patients to chronic sub-acute ischemia.