Child's Nervous System

, Volume 36, Issue 1, pp 59–71 | Cite as

Value of computerized shunt infusion study in assessment of pediatric hydrocephalus shunt function—a two center cross-sectional study

  • Sandra Fernandes Dias
  • Afroditi–Despina Lalou
  • Regine Spang
  • Karin Haas-Lude
  • Matthew Garnett
  • Helen Fernandez
  • Marek Czosnyka
  • Martin U. SchuhmannEmail author
  • Zofia Czosnyka
Focus Session



Hydrocephalus shunt malfunction can—also in children—occur insidiously without clear symptoms of raised intracranial pressure (ICP) or changes in ventricular size, imposing a diagnostic challenge. Computerized shunt infusion studies enable quantitative shunt function assessment. We report on feasibility and results of this technique in children in a two center cross-sectional study.

Material and methods

Shunt infusion study (SIS) is performed with two needles inserted into a pre-chamber for ICP recording and CSF infusion. After baseline ICP recording, constant rate infusion is started until a new ICP plateau (ICPpl) is reached. Dedicated software containing the shunt’s resistance characteristics calculates ICP and its amplitude outflow resistance and critical shunt pressure (CSP). Overall, 203 SIS were performed in 166 children. Shunts were defined as functional if ICPpl was <CSP and obstructed if ICPpl was > 5 mmHg above CSP and borderline in between.


Forty-one shunts (20.2%) were found obstructed, 26 (12.8%) had borderline characteristics, and 136 (67%) were functional. Baseline ICP in obstructed shunts was significantly above shunt operating pressure. CSF outflow resistance (Rout) and ∆ICP plateau were significantly elevated in obstructed shunts, with cut-off thresholds of 8.07 mmHg min/ml and 11.74 mmHg respectively. Subgroup analysis showed smaller ventricles in 69% of revised cases.


SIS is a feasible, reliable, and radiation-free technique for quantitative shunt assessment to rule out or prove shunt malfunction. Dedicated software containing shunt hydrodynamic characteristics is necessary and small children may need short-term sedation. Due to the clinical and inherent economic advantages, SIS should be more frequently used in pediatric neurosurgery.


Hydrocephalus Shunt infusion study Pediatric neurosurgery Intracranial pressure Shunt dysfunction 



Fundamental amplitude (1st harmonic) of ICP after Fourier transformation


Cerebro spinal fluid (Cerebrospinal fluid)


Critical shunt pressure determined by software according to shunt characteristics


Evans’ index


Fronto-ccipital horn ratio


Intracranial pressure


ICP at plateau phase of infusion study


Magnetic resonance imaging


Expected intracranial pressure at the operating pressure of the shunt valve


Resistance to CSF outflow


Ventriculo-peritoneal shunt

ΔICP (mmHg)

Difference between ICP at baseline and ICP plateau

ΔAMP (mmHg

Difference between AMP at baseline and AMP at plateau

ΔICPop (mmHg)

Difference between theoretical operating pressure according to valve setting and ICP at baseline

ΔCSP (mmHg)

Difference between ICP at plateau and CSP

ΔRout (mmHg min/ml)

Difference between measured Rout and laboratory-tested Shunt Rout


Accurate diagnosis of shunt malfunction is a clinical problem presenting many facets and challenges. Clinical presentation in children can involve clear symptomatology of raised intracranial pressure (ICP) [19, 22, 24, 37]. When symptoms are accompanied by ventricular enlargement, there is no question about the necessity of shunt revision. However, shunt malfunction can also present with disputable and less specific symptomatology and/or unchanged ventricles. Some children, although clinically not obviously affected, might not show the expected decrease of ventricular size and widening of external CSF spaces after shunt implantation, despite a low or lowered shunt-opening pressure [6, 18, 28]. Thus, there exist many dilemmas in everyday neurosurgical practice. When objectively testing, shunt puncture and snapshot pressure measurement is of little value in case of pressure compensated (normal pressure) shunt dysfunction (usually with chronic adaptations) or might be influenced by circumstances of measurement like agitated child. Radiographic studies looking a dye clearance are at best of qualitative nature and carry risks of radiation exposure in small children or induction of shunt obstruction by viscous radio-opaque dyes [6, 10, 13, 22, 24, 19, 35]. Moreover, in cases when the clearance is sluggish, because of low pressure at the time of measurement, false results may be achieved. It has been reported in the adult population that shunt testing in vivo allows management of patients effectively, without revising a properly functioning shunt (or one which just requires a setting change) [7, 10, 12, 13, 28, 31].

Open shunt revision with intra-operative shunt testing is very invasive. Particularly, in the pediatric population, it has been demonstrated that multiple shunt revisions are a prognostic factor for worse cognitive development, further shunt revisions, and shunt dependency [2, 29, 32, 36]. In case of an intra-operative functioning shunt, an unnecessary procedure with significant risks of shunt infection had been undertaken. Even iatrogenic disruption of a functioning shunt can occur [36].

On the other hand, observation instead of action can be an element of delay in treatment, or long-term neglect, especially in patients with complex postural over-drainage or intermittent obstruction [17, 25, 29, 32].

Finally, from a financial point of view, the adverse events of infections, (unnecessary) multiple surgical revisions, and poor developmental outcomes can be translated to a significant waste of funds for national healthcare systems as well as significant burden to families and societies [2, 17, 23, 29, 32, 41, 42].

Shunt testing in vivo using shunt infusion studies (SISs) has been implemented in clinical practice and reported in the literature for over 25 years [13, 21, 30, 31, 42]. It is an accurate and relatively low-invasive method to assess shunt function and detect over-drainage and underdrainage as well as proximal/distal blockage. Previous studies have shown the efficacy of SIS in the adult population with implanted ventricular-peritoneal shunts for hydrocephalus after subarachnoid hemorrhage or in normal pressure hydrocephalus [1, 4, 7, 11, 13]. The established safety of this technique allows it to be used in children.

This study investigates the value of SIS as a tool to support clinical decisions in the pediatric hydrocephalus population. We retrospectively analyzed SIS data performed in children in two centers with long-term experience in shunt infusion studies.

Patients and methods

Patient cohort

The presented study retrospectively included all pediatric patients aged under 16 years old that were submitted to a shunt infusion study (SIS) as part of the clinical routine workup for suspected shunt malfunction. The inclusion period encompassed January 2003 to August 2017, with the SISs being performed in either the Department of Neurosurgery Addenbrooke’s Hospital in Cambridge (UK) or in the Department of Neurosurgery at the University Hospital of Tübingen (Germany). The retrospective data collection and analysis was approved by the institutional ethics review board of Tübingen (Ref. 160/2018BO2). From Cambridge Institution, no additional ethical approval was required, as SIS is a clinical tool that has been performed after clinical request. Before intervention, parents were informed about the procedure and possible related complications (e.g., infection) and gave a consent.

Selection criteria for SIS varied between consultants. In general, patients presenting with mild or subtle symptoms of shunt malfunction and incoherent imaging (ventricles unchanged or smaller) or contradictory findings of symptoms and imaging (like no change in symptoms but ventricles mildly increased or ventricles remained large despite significant opening of adjustable shunts) were usually selected for a SIS. Children with obvious clinical symptoms and signs of raised intracranial pressure and shunt malfunction—headaches, vomiting, epileptic seizures, papilledema, increasing macrocephaly—and/or radiological features (e.g., enlarged ventricles) were not included (Fig. 1).
Fig. 1

Flow diagram on the selection and categorization of pediatric patients’ shunt infusion studies. Group A included patients who did not require revision and were still well on follow-up. Group B included patients whose shunt function appeared questionable and who required a shunt adjustment, but the shunt was not blocked in short-term follow-up or revision. Group C included patients who underwent revision soon after the infusion and the shunt was confirmed to be obstructed intraoperative

Sedation and anesthesia management

In Tübingen, children who are unable to cooperate with the procedure awake (below the age of 7–8 years or mentally impaired) received mild sedation with propofol (initial bolus of 2–4 mg/kg) at a rate of 1–4 mg/kg/h. Whenever feasible, studies were performed in awake children without any medication. Listening to music or audiobook was implemented for distraction and child well-being during the procedure.

In Cambridge, children that were uncooperative or had severe disabilities were studied under general anesthesia. Occasionally, mild sedation with chloral hydrate or nitrous oxide was given. In case it was feasible, children were studied fully conscious with play activities at the bedside.

SIS technique

The protocol for performing SIS has been previously described [12, 38]. The method is based on a continuous ICP measurement and infusion of a sterile mock CSF solution into the most proximal tap reservoir. The tap reservoir could be between ventricular catheter and valve (the most common situation) or be an independent reservoir access to the ventricular system (under rare circumstances of complicated hydrocephalus).

Both institutions performed SIS under the same operating standards. Before starting, a safe-checklist is performed, assuring patient’s data (which at this point is inserted on the ICM+ software), medical indication, and consent from the parents. Thereafter, SIS starts, encompassing the following steps (see Fig. 2):
Fig. 2

Shunt infusion study settings. a Under sterile conditions, two hypodermic (25 gauge) needles are inserted at the shunt pre-chamber. b One needle is connected to a manometer line and a pressure transducer. c Through the other needle, the sterile solution is infused at a constant rate from an infusion-pump. d A laptop computer harboring the ICM+ software records and displays parameters as ICP, AMP, among others

Under sterile conditions, manometer lines and transducers are carefully filled with Ringer’s or Hartman solution.
  1. 1.

    After thorough skin disinfection, under aseptic conditions, a surgical drape with a hole is placed over the tapping reservoir (Fig. 2a).

  2. 2.

    Two hypodermic needles (gauge 25) are inserted into shunt pre-chamber (Fig. 2a), one for the infusion line connected to the infusion pump (Fig. 2c), the other for to the manometer line connected to a pressure transducer (Fig. 2b).

  3. 3.

    The pressure transducer (± pressure amplifier) is connected via an AD converter to a laptop computer running ICM+ software (Fig. 2d)

  4. 4.

    A baseline ICP recording of about 10 min is performed.

  5. 5.

    Thereafter, infusion starts at a rate of 1 ml/min or 1.5 ml/min

  6. 6.

    The ICP rise is observed on the monitor until a stable plateau is reached and maintained for about 10 min. Then, infusion is stopped and ICP decrease is observed for a few minutes before needles are removed.


During the entire SIS, children are clinically monitored by qualified medical staff that assure safeness of the procedure. No prophylactic antibiotics are administered for performing SIS.

Data collection and analysis with ICM+ software

ICM+ software has been introduced into the Brain Physics Laboratory in Cambridge for more than 15 years [38, 39]. The hydrodynamic properties of all shunt systems tested (with a non-shortened distal catheter connected) are incorporated into the ICM+ software. The continuous database expansion of shunt systems tested in the Cambridge Laboratory allows the clinical assessment of shunt function of the great majority of the shunt devices used nowadays. For all shunt types tested in this study, hydrodynamic properties were known to the software. Thus, a critical shunt pressure (CSP) could be calculated, which—given a normal resistance of shunt and distal catheter—should not been surpassed at a given shunt infusion rate.

The ICP trace is recorded with a sampling frequency of 100 Hz and every 10 s, a mean value of ICP is calculated, which forms the basis for all further ICP-based calculations. After concluding SIS, the time of baseline and plateau ICP recording is manually defined, as well as the transition zone of rising ICP in between and finally the total time of infusion. The shunt valve and its pressure settings are selected and a “best-fit curve” is calculated (see Fig. 3). The software furthermore computes the following parameters: ICP and AMP at baseline and plateau, the critical shunt pressure (CSP), the outflow resistance of the whole system (Rout), and the elasticity (E) according to Marmarou model [5, 9, 27].
Fig. 3

Shunt infusion study of a functional shunt. a CT-scan of a 4-year-old child presenting with enlarged ventricles after VP-shunt placement and valve replacement for a ProGav 4/29 (initial Hackim Medos 10) due to mal-resorption hydrocephalus after AVM bleeding. Clinically, she was stable with slow recovery and had no complains. b SIS revealing a gradual increase of the ICP during the infusion, with the plateau being reached below the shunt critical pressure (upper panel); amplitude was below 1 mmHg (panel below)—indicating a normal functioning shunt with adequate brain compliance. No VP shunt revision was performed and the child kept improving its clinical condition and development

We furthermore determined the so-called operating pressure of the individual shunt system according to its pressure setting (ICPop) [14, 16]. The ICPop ± a tolerance zone of max 5 mmHg in adults (to compensate for influence of abdominal pressures) is the pressure range to which a functioning shunt should regulate the ICP at rest at a presumed CSF production rate of 0.3 ml/min. The difference of the recorded baseline ICP to the ICPop was determined (∆ICPop).

The amount of ICP and AMP increase between baseline ICP and plateau ICP (∆ICP and ∆AMP) was calculated. The short duration of the SIS and strong non-linear influence of the shunt system does not allow for a reliable interpretation of elasticity/compliance values E and PVI [15]; therefore, this parameter was not used for the purpose of the present study.

Shunt function classification

The shunts were classified as functional when the infusion study appeared normal—with the ICP plateau being below the critical shunt pressure. An adjustment of the setting was required in some cases and it helped improve patient’s symptoms (group A, see Fig. 3).

Given the fact that the intra-abdominal pressure usually is in the range of max 5 mmHg, we took a cut-off of ≥ + 5 mmHg for the ICP plateau above the CSP to classify a shunt as obstructed or non-functional. Patients in which the SIS identified an obstructed shunt were all surgically revised (group C, see Fig. 4).
Fig. 4

Shunt infusion study of a non-functional shunt. a Head-circumference graphic of an initially 5-month-old child when presenting with macrocephaly and diagnose of a Blake’s pouch (b). c Four years after VP shunt implantation, despite normalization of the head circumference and “pumping" shunt prechamber the ventricles remained enlarged and the child presented with developmental delay. d SIS showed a significant increase of the ICP, reaching an ICP plateau pressure of 35 mmHg clearly above the critical shunt pressure (15 mmHg), this way diagnosing shunt malfunction. The shunt was revised and during surgery, a kink of the connecting catheter between the burr-hole reservoir and valve was found which was invisible on the pre-operative skull X-ray. That explained the largely increased shunt resistance demonstrated by SIS

In case SIS showed an ICPplateau being ≤ 5 mmHg above the critical shunt pressure, the shunt function was classified as borderline. Taking all clinical and radiological information together, the responsible consultant decided on shunt revision or observation ± shunt adjustment (group B).


The clinical and radiological follow-up was performed according to clinical standards set by the responsible consultants in both institutions.

The subset of the children investigated with SIS in Tübingen were analyzed in detail regarding their radiological follow-up after SIS. Furthermore, all children received at least one clinical follow-up in outpatient clinics at 3–6 months after SIS according to the individual situation. In case of shunt revision, a MRI scan was usually obtained 6 months later. Clinical notes from the first two visits after SIS were analyzed according to a change in clinical status (unchanged, improved, and worsened). The ventricular size in the last imaging before SIS and the first thereafter was compared using Evans’ index (EI) and fronto–occipital horn ratio (FOHR).

Statistical analysis

All data were processed using R version 3.3.3. We checked for normal distribution. Quantitative variables are presented as mean ± standard deviation. The Wilcoxon test was used to compare differences between the different groups and ROC and multiroc analysis was performed using the pROC package [33].


A total number of 203 SIS (164 in Cambridge and 39 in Tübingen) were performed in 166 children (see Fig. 1 for selection and categorization of SIS). At the time of the SIS, children’s mean age was 8.6 ± 5 years (min 1 month; max 16 years) with 52% (n = 86) of the children being female and 48% (n = 80) male. In 23 children, 2 SISs were performed and ≥ 3 SISs were performed in 14 children.

For detail on distribution of shunt valves according to each institution refer to Table 1.
Table 1

Shunt valve distribution according to institution

Shunt valve




CSF flow control valve


  Burr hole median

  Contoured median






Delta Valve




Medos Hakim Progr. Valve




Orbis Sigma
















Sophysa Polaris








Strata NSC








Functional shunts (group A)

In 136 (67%) SIS, the ICPplateau was below the critical shunt pressure (CSP) with a mean difference to critical shunt pressure (∆CSP) of − 4.68 ± 4.07 mmHg and the shunt was rated as functional, also on follow-up. ICPbaseline was close to the expected operating ICP (ICPop); thus, ∆ICPop (deviation of baseline to operating ICP) was small with − 1.15 ± 3.88 mmHg. ∆ICP (the difference between ICPbaseline and ICPplateau) was 6.19 ± 3.35 mmHg.

Mean pulse amplitude of intracranial pressure waveform at baseline was 0.49 ± 0.39 mmHg. CSF outflow resistance was 4.42 ± 2.0 mmHg min/ml.

Figure 3 illustrates a clinical case of a functional shunt with the respective curves obtained during SIS.

Borderline shunts (group B)

Twenty-six (12.8%) SISs had borderline function with ICPplateau ≤ 5 mmHg above the critical shunt pressure, resulting in average ∆CSP of 3.73 ± 2.95 mmHg.

ICPbaseline was above that of group A and at the upper limits of the tolerance zone for ICPop; thus, ∆ICPop was 4.77 ± 3.87 mmHg. Both values were significantly different from group A (see Table 2).
Table 2

Results of cerebrospinal fluid dynamics as derived from the shunt infusion study in the 3 different groups. CSP, critical shunt pressure; Rout, resistance to CSF outflow; ICP, intracranial pressure; AMP, fundamental amplitude of ICP; ΔICP, ICP plateau - ICP baseline; ΔAMP, AMP plateau - AMP baseline; ΔICPop, ICP baseline-shunt operating ICP; ΔCSP, ICP plateau - shunt CSP. Statistically significant p values are displayed in italics


Functioning (group A, N = 136)

Borderline (group B, N = 26)

p value (groups A and B)

Blocked (group C; N = 41)

p value (groups B and C)

p value (groups A and C)

Operating pressure (mmHg)

7.22 ± 2.07

6.95 ± 2.20


6.65 ± 2.33



CSP (mmHg)

16.94 ± 2.77

16.67 ± 2.34


15.45 ± 3.18



Shunt Rout (mmHg min/ml)

3.20 ± 0.81

3.49 ± 0.82


3.49 ± 0.93



ICPbaseline (mmHg)

6.07 ± 3.85

11.71 ± 4.47


11.41 ± 5.89



ICPplateau (mmHg)

12.26 ± 4.31

20.40 ± 3.67


29.42 ± 10.22



AMPbaseline (mmHg)

0.49 ± 0.39

0.68 ± 0.47


0.93 ± 0.61



AMPplateau (mmHg)

0.79 ± 0.62

1.17 ± 0.82


2.76 ± 2.04



Rout (mmHg min/ml)

4.42 ± 2.00

6.80 ± 2.67


14.82 ± 6.39



ΔICP (mmHg)

6.19 ± 3.35

8.69 ± 3.29


18 ± 8



ΔAMP (mmHg)

0.3 ± 0.4

0.49 ± 0.56


1.83 ± 1.8



ΔICPop (mmHg)

− 1.15 ± 3.88

4.77 ± 3.87


4.76 ± 5.82



ΔCSP (mmHg)

− 4.68 ± 4.07

3.73 ± 2.95


13.97 ± 9.49



∆ICP (ICPplateau minus ICPbaseline) was 8.69 ± 3.29 mmHg. AMPbaseline was 0.68 ± 0.47 mmHg. Mean Rout was 6.80 ± 2.67 mmHg min/ml. All parameters were significantly elevated compared with group A (see Table 2).

In this group, one child underwent shunt revision since the responsible consultant pediatric neurosurgeon decided on revision. In this case, the shunt was found not obstructed, and the child improved clinically on follow-up.

Obstructed/blocked shunts (group C)

Forty-two SIS (20.7%) were classified as obstructed with ICPplateau above CSP on average by 13.97 ± 9.49 mmHg (statistically significant greater than in groups A and B, see Table 2). ICPbaseline was not different to group B, but significantly greater than in group A.

In rare cases of occlusion, there was no ICP plateau, as the pressure continued to rise in a linear fashion. In these cases, infusion was terminated whenever ICP crossed 35 mmHg and ICPplateau taken as 35 mmHg.

ICPbaseline in this group was significantly above ICPop by average ∆ICPop of 4.76 ± 5.82 mmHg, significantly elevated in comparison with group A, not B.

AMPplateau was 2.76 ± 2.04 mmHg. Increase in ICP during infusion was 18 ± 8 mmHg, and Rout 14.82 ± 6.39 mmHg min/ml. All values were significantly elevated compared with groups A and B (Table 2).

In all children of this group, a shunt obstruction was found at surgical revision, either of the valve or the distal catheter, with replacement of the respective non-functional part. Figure 4 illustrates a case with non-functional shunt.

Using the cohort of functioning shunts (A) versus those of borderline and obstructed shunts (B and C), we calculated cut-off values for ∆ICP, AMPplateau, and Rout, that could, in addition to a ICPplateau−CSP of > 5 mmHg, predict shunt obstruction. These values are potentially independent of the knowledge of the physical characteristics of the individual shunt system. Cut-off values were determined using ROC analysis and were 11.74 mmHg, 1.25 mmHg, and 8.07 mmHg min/mL, respectively. Excellent discriminatory abilities with high statistic significance were found, see Figs. 5 and 6.
Fig. 5

SIS results analysis between functional (group A) and non-functional (groups B and C) shunts. a Delta operating ICP: ICPbaseline - shunt operating pressure. b Delta ICP: ICPplateau.- ICPbaselinecRout: resistance to CSF outflow

Fig. 6

Receiver operating characteristics curves for Rout (left panel), delta ICP (middle), and AMP plateau (right panel) between functional and non-functional shunts. The respective cut-off values, so as sensitivity, specificity, negative predictive value (NPV), and area under the curve (AUC) are given within the graphics

Supplementary Tübingen subgroup analysis

In Tübingen, 30 children with suspected asymptomatic or oligo-symptomatic underdrainage were submitted to 39 SIS. Of those children, radiological follow-up data was analyzed in detail.

Clinical symptoms

All 16 children with functional shunts at 19 SIS remained stable or improved their neurological abilities according to their expected natural development during follow-up time, which was 33 ± 24 months. One child had SIS indicating shunt obstruction at a later time point, after some new but inconclusive symptoms appeared.

Two of 3 children following a borderline SIS improved clinically after shunt adjustment (lower opening pressures) within a short time, and the remaining child was clinically stable (SIS revealed low reserve/compliance).

Ten of the 14 children (71%) with obstructed shunts that all underwent shunt revision improved in their clinical condition and neurological development at 6 months in the perception of their parents and the impression of the pediatric neurologists, while the remaining 4 children remained unchanged according to their previous status.

Radiological features

Radiological follow-up by magnetic resonance imaging (MRI) was available in 37 of 39 SIS events (95%), within a mean follow-up of 37 ± 29 months after SIS.

In the functional group A, 10 out of 16 children (62%) had at last follow-up smaller and 5 unchanged ventricles (one child did not show for follow-up), with ∆FOHR and ∆EI (between last follow-up and pre-SIS) being 0.046 ± 0.08 and 0.037 ± 0.07 respectively. Changes were statistically significant comparing last MRI with before SIS.

The 14 children of group C were submitted to 17 SIS and respective surgical revision due to shunt obstruction (3 children had 2 obstructions at different time points). Nine of 14 (64%) showed on last MRI follow-up after last revision smaller ventricles, and 5 children (31%) presented with unchanged ventricles. The ∆FOHR was 0.049 ± 0.097, and ∆EI 0.043 ± 0.097. Ventricle changes within this group were statistically significant to last MRI before SIS.

All radiological features and respective statistical analysis are summarized in Table 3.
Table 3

Results of ventricle measurements on MRI of the 3 different groups in subgroup analysis from Tübingen (N = 39). FOHR, frontal and occipital horn ratio; EI, Evan’s index; Δ, follow-up measurements–pre-infusion measurements. p value, difference between pre-SIS and follow-up within group


Functioning (group A; N = 19)

p value (A)

Borderline (group B; N = 4)

p value (B)

Blocked (group C; N = 16)

p value (C)

Follow-up MRI (months)

36.5 ± 26.19


50.33 ± 37.65


39.3 ± 33.8


FOHR pre-infusion

0.49 ± 0.1


0.5 ± 0.11


0.46 ± 0.08


FOHR follow-up

0.45 ± 0.1


0.43 ± 0.059


0.42 ± 0.062



0.046 ± 0.08


0.073 ± 0.087


0.049 ± 0.097


Evans’ pre-infusion

0.37 ± 0.1


0.34 ± 0.11


0.38 ± 0.084


Evans’ follow-up

0.34 ± 0.1


0.3 ± 0.07


0.33 ± 0.079


ΔEvans’ index

0.037 ± 0.07


0.047 ± 0.063


0.043 ± 0.097



In this series of patients, no complications related to the procedure were identified. In 0 of 166 SIS, a shunt infection was found within 3 months of the procedure.

No clinical symptoms of increased intracranial pressure during the elevated plateau pressure during infusion in case of under-draining shunts were observed.


Shunt dysfunction in children

Shunt malfunction in children can occur with clear symptomatology of raised intracranial pressure. In such a case plus some degree of enlargement of ventricular size, there is no question about the necessity of shunt revision. However, shunt dysfunction can also present with disputable and less specific symptoms and unchanged ventricles, or children with unchanged clinical status show mildly enlarging ventricles at unchanged shunt setting in routine imaging. Others, although clinically not obviously affected, might not show the expected decrease of ventricular size and widening of external CSF spaces after shunt implantation or revision, despite a low or lowered shunt-opening pressure. In these scenarios remains too much uncertainty to submit a child to surgery just on the basis of suspicion. This report shows that in case of suspected underdrainage, a SIS is the most precise, since quantitative, examination to answer the question of an underlying shunt malfunction.

Proximal catheter assessment prior to infusion

The patency of the proximal catheter is assessed by recording a normal ICP pressure wave form during baseline recording with a peak-to peak wave amplitude of at least 1 mmHg in infants. Peak-to-peak ICP amplitudes with a functioning shunt system are to be expected to be higher between 2 and 4 mmHg, which translates to an AMP value of approximately 0.5–1 mmHg. Absent or low peak-to-peak ICP amplitudes (< 1 mmHg) or AMP < 0.25–0.5 are highly suspicious of either partial proximal catheter obstruction or occlusion, especially if there is no immediate ICP increase on coughing or breath holding (or jugular compression in a sedated child). Alternatively, a wrong needle placement outside the tap reservoir has to be suspected and ruled out.

In all three groups, mean AMPbaseline was accordingly above this threshold between 0.49 and 0.93 mmHg, proving open ventricular catheters.

In contrast to pumping, an amplitude determination provides an objective picture. It also has less risks than repetitive pumping, which may (rarely) produce acute intracranial hypotension, with all possible adverse consequences [37].

Discrimination of valve and distal catheter function

As soon as the infusion is started, the whole system distal to the insertion point of the two needles is assessed as a whole. Any pathological resistance in the system will lead to an ICP plateau above the CSP. Therefore, SIS cannot differentiate between an obstruction within the valve or of the abdominal catheter, or in both. Consequently, both components have to be tested individually at surgical shunt revision with water column–based run-off tests.

Influence of shunt valves and indication for shunt infusion study

In this study, 10 different valve types were investigated. Infusion rates were performed at both 1 ml and 1.5 ml per min and all critical shunt pressures were according to values previously calculated in laboratory settings [12]. In Cambridge, there was a preference for Strata valves, which can have a Delta chamber for over-drainage protection. The clinical question leading to SIS has also been over-drainage in many cases, which can be answered by a sit-up test of the patient with the pressure monitoring still in place. In case of over-drainage, the ICP will not only initially and physiologically decrease due to the outflow of CSF into the spinal canal and accelerated blood outflow from the head, but also continue to fall over the next 20–30 min due to continuous accelerated CSF drainage through the shunt. The SIS however will show normal values. In Tübingen, the clinical question had exclusively been underdrainge. This difference in indication explains the relatively larger proportion of functional shunts in the Cambridge population as compared with the Tübingen cohort (71% versus 49%) and the lower rate of shunt revision (15.2% versus 46%).

ICM+ software-based analysis

Clinical interpretation of all variables derived from SIS is an important aspect to make further use of the provided information than just looking at critical shunt pressure and its relation to the plateau pressure. ICP at baseline and plateau, AMP at baseline and plateau, and total Rout allow the establishment of differential diagnoses in clinical overlapping entities [20].

Since a functioning shunt should maintain the ICPbaseline around its opening pressure, we calculated in each patient the “expected ICP at baseline” according to the known shunt characteristics at a presumed normal CSF production rate of 0.3 ml/min, the so-called operating pressure or ICPop. To compensate for influence of the abdominal pressure (normally around 5 mmHg in non-obese children), we defined a corridor of ± 5 mmHg as the normal limits of ICPop in a functioning shunt. The mean ICPbaseline in functional shunts was in the middle of this corridor, and borderline shunts and obstructed shunts were at the upper border of this 5-mmHg corridor. This indicates that shunts defined as “borderline” indeed seem to have some compromise in CSF drainage which leads to an increased ICPbaseline in those patients.

Intracranial pulsatility, best expressed by AMPbaseline, was lowest in the functional group, indicating that the intracranial compliance was normal and the intracranial compartment physiologically relaxed. In borderline shunts, there was a significant increase in AMPbaseline with a further significant increase in obstructed shunts. This indicates that dysfunctional shunts with a higher resistance not only regulate the intracranial pressure at higher levels, but that there is in addition a significant decrease of intracranial compliance with increased intracranial pulsatility as a measurable physical hallmark of insidious shunt dysfunction. Notably, while ICPbaseline and its difference from ICPop were not different between borderline and obstructed shunts, AMPbaseline as a marker of compliance in the resting state was significantly higher in the obstructed shunts, indicating that compliance is more sensitive to insidious shunt underdrainage than pressure.

Further parameters and cutoffs to indicate shunt obstruction

This study includes many different shunt valves with different resistances and pressure settings at the time of testing. The software knowing the characteristics of the investigated shunt is therefore key for quantitated results. However, this study also identifies ∆ICP, AMPplateau, and Rout, as significantly different between functional and dysfunctional shunts and these parameters are independent of the knowledge of the shunt characteristics or the knowledge of the shunts operating pressure. For these parameters, cut-off values could be calculated which identified a shunt that is obstructed with high sensitivity and specificity (see Fig. 6). This means, if i∆ICP, AMPplateau, and Rout are above their threshold levels of 11.74 mmHg,1.255 mmHg, and 8.07 mmHg min/mL, respectively, then there is a high possibility that the shunt belongs to the dysfunctional group according to the precise knowledge-based definition of being more than 5 mmHg above critical shunt pressure.

This is of importance because it might allow a shunt to be tested with a much simpler shunt infusion technique, by just recording ICP and deriving, e.g., ∆ICP and Rout from the ICP trace without sophisticated software. Certainly, it has to been shown in further research that these thresholds calculated from 10 different types of valves are reliable for other shunts valves too, but for the moment, this might serve as a first indicator of shunt dysfunction. This might be helpful for neurosurgical services in health systems with a lower economic status that cannot afford computers and software, but still can measure and record ICP and thus can perform a simplified shunt infusion study. Further research is necessary to investigate if an extrapolation of these results has any meaning.

Tübingen subgroup analysis

All children in Tübingen (n = 30) underwent on 39 occasions, a SIS for suspicion of silent underdrainage or blockage. This means we investigated mildly or asymptomatic children with incongruence of imaging and clinical presentation. Shunt malfunctions of different degrees (group B borderline and group C obstructed) occurred in 51% of all SIS.

All of group C (14 of 30 children, 17 SIS) underwent shunt revision, and 10 of 14 (71%) improved clinically in the perceptions of parents and pediatricians; however, no objective assessment by a formal neuro-developmental testing was performed pre- and post-operatively. Therefore, we cannot prove that identification of a blocked shunt and re-installing a functional shunt did objectively change the clinical course of those patients. However, a non-functional shunt in a child with hydrocephalus makes little sense, even if it currently has no obvious symptoms of pressure active hydrocephalus.

Regarding the ventricular size, 9 of 14 children (64%) showed a significant decrease of ventricles after revision and this way exhibited the same behavior than those children with functional shunts, in whom 10 of 16 (62%) decreased in a comparable follow-up time with a mean of about 3 years. Therefore, from a neuroradiological point of view, a normalization of intracranial conditions could be achieved by shunt revision. In summary, this demonstrates that children with insidious or silent shunt malfunction that cannot be diagnosed without SIS have a radiological profit from the shunt revision and most likely in the long term also from the clinical point of view.

Clinical safeness

No complications related to the SIS were observed in this study, namely no infections. An internal audit performed between 2007 and 2009 (unpublished results) at the Department of Neurosurgery in Cambridge revealed a low rate < 1% of CSF infection following all SISs (including the more frequent SIS in adults) if care is taken during the entire procedure—in particular what concerns skin disinfection and filling of the manometer lines and transducer [42].

This data demonstrates a much lower complications rate of SIS when compared with surgical shunt revision, where a shunt infection rate of 2–38% has been reported [18, 20, 40]. Given the extremely low infection rate of SIS in children (0% in this sample), this furthermore encourages the use of SIS in case of questionable shunt function as opposed to open surgical shunt testing. Finally, it could be shown that multiple shunt revisions not only relate to worse neurocognitive development in children, but represent an independent risk factor for further revisions as well [16, 17].

Health economic advantages

It has been calculated that performing shunt function assessment through SIS in comparison with invasive surgical revision comports economic advantages [22]. The usual costs associated with operative shunt function assessment like cost of inpatient treatment for at least 1–3 days and operating room time clearly contrasts to the minimal costs for single-use material needed to perform a shunt infusion study in the context of an outpatient setting, even if a short-term sedation is required.

Alternative methods

Other methods of in vivo shunt function assessment involve radiographic dye clearance tests, thermo-dilution methods, or sonography-based detection of shunt flow [3, 8, 18, 26, 34, 43]. The main limitation of all of this methods is that the driving pressure is unknown as well as the resistance of a functioning shunt system and therefore all are qualitative tests and do only allow—at best—to detect a blocked shunt. A partially obstructed shunt with an increased internal resistance, as we found in the majority of all dysfunctional systems, however cannot be reliably identified with this mentioned methods. This is one reason among others why semi-quantitative methods have not made it into clinical routine of shunt work-up in the vast majority of neurosurgical units.


One limitation of the present study is the retrospective character. The fact that the data were collected from two different centers does not allow a complete homogenization of inclusion criteria for performing the SIS, as this depends on standards of clinical practice implemented in each institution. This can be interpreted as selection bias, which might have led to a higher percentage of patients with functional shunts in Cambridge.

Regarding practical application, children that are not cooperative do need to be sedated or studied under general anesthesia. This certainly brings in some additional “invasiveness”; however, this still seems to be less invasive when compared with open shunt revision.

In order to be able to perform SIS, the department must acquire ICM+ software (Cambridge Enterprise Ltd) or built its own software tools and one or more members have to be trained with the technique—which has however a relatively fast learning curve. The economic benefits associated with SIS overwhelm rapidly the initial costs of the software.


Shunt infusion study is a feasible, elegant, and radiation-free technique for quantitative shunt assessment in children of all ages to rule out or prove insidious shunt malfunction. Since silent shunt dysfunction is nothing else but another form of compensated—not sufficiently treated—chronic hydrocephalus, with all possible consequences to the development of the child, SIS is an important tool for the pediatric neurosurgeon. It provides an assurance that children with hydrocephalus under care truly have functional shunts or revision is necessary. A clinical benefit accompanied by radiologic improvement is clearly demonstrated by the subgroup analysis of the Tübingen cohort.

Due to the clinical and inherent economic advantages, we postulate that SIS should become routine in neurosurgery units.



MC is supported by the National Institute of Health Research, Cambridge Centre.

Compliance with ethical standards

The retrospective data collection and analysis was approved by the institutional ethics review board of Tübingen (Ref. 160/2018BO2). From Cambridge Institution, no additional ethical approval was required, as SIS is a clinical tool that has been performed after clinical request. Before intervention, parents were informed about the procedure and possible related complications (e.g., infection) and gave a consent.

Conflict of interest

MC receives part of licensing fees for the software ICM+used for data collection and analysis in this study. The other authors have no conflict of interest to report


  1. 1.
    Andersson N, Malm J, Backlund T, Eklund A (2005) Assessment of cerebrospinal fluid outflow conductance using constant-pressure infusion--a method with real time estimation of reliability. Physiol Meas 26:1137–1148. CrossRefPubMedGoogle Scholar
  2. 2.
    Arrington CN, Ware AL, Ahmed Y, Kulesz PA, Dennis M, Fletcher JM (2016) Are shunt revisions associated with IQ in congenital hydrocephalus? A meta-analysis. Neuropsychol Rev 26:329–339. CrossRefPubMedGoogle Scholar
  3. 3.
    Benzel EC, Mirfakhraee M, Hadden TA (1991) Evaluation of CSF shunt function: value of functional examination with contrast material. AJNR Am J Neuroradiol 12:143–147PubMedGoogle Scholar
  4. 4.
    Borgesen SE, Albeck MJ, Gjerris F, Czosnyka M, Laniewski P (1992) Computerized infusion test compared to steady pressure constant infusion test in measurement of resistance to CSF outflow. Acta Neurochir 119:12–16CrossRefGoogle Scholar
  5. 5.
    Bouma GJ, Muizelaar JP, Bandoh K, Marmarou A (1992) Blood pressure and intracranial pressure-volume dynamics in severe head injury: relationship with cerebral blood flow. J Neurosurg 77:15–19. CrossRefPubMedGoogle Scholar
  6. 6.
    Boyle TP, Nigrovic LE (2015) Radiographic evaluation of pediatric cerebrospinal fluid shunt malfunction in the emergency setting. Pediatr Emerg Care 31:435–440; quiz 441-433. CrossRefPubMedGoogle Scholar
  7. 7.
    Chari A, Czosnyka M, Richards HK, Pickard JD, Czosnyka ZH (2014) Hydrocephalus shunt technology: 20 years of experience from the Cambridge Shunt Evaluation Laboratory. J Neurosurg 120:697–707. CrossRefPubMedGoogle Scholar
  8. 8.
    Chiewvit S, Nuntaaree S, Kanchaanapiboon P, Chiewvit P (2014) Assessment lumboperitoneal or ventriculoperitoneal shunt patency by radionuclide technique: a review experience cases. World J Nucl Med 13:75–84. CrossRefPubMedPubMedCentralGoogle Scholar
  9. 9.
    Czosnyka M, Czosnyka Z, Agarwal-Harding KJ, Pickard JD (2012) Modeling of CSF dynamics: legacy of Professor Anthony Marmarou. Acta Neurochir Suppl 113:9–14. CrossRefPubMedGoogle Scholar
  10. 10.
    Czosnyka M, Czosnyka Z, Momjian S, Pickard JD (2004) Cerebrospinal fluid dynamics. Physiol Meas 25:R51–R76CrossRefGoogle Scholar
  11. 11.
    Czosnyka M, Pickard JD (2004) Monitoring and interpretation of intracranial pressure. J Neurol Neurosurg Psychiatry 75:813–821CrossRefGoogle Scholar
  12. 12.
    Czosnyka M, Whitehouse H, Smielewski P, Simac S, Pickard JD (1996) Testing of cerebrospinal compensatory reserve in shunted and non-shunted patients: a guide to interpretation based on an observational study. J Neurol Neurosurg Psychiatry 60:549–558CrossRefGoogle Scholar
  13. 13.
    Czosnyka Z, Czosnyka M, Owler B, Momjian S, Kasprowicz M, Schmidt EA, Smielewski P, Pickard JD (2005) Clinical testing of CSF circulation in hydrocephalus. Acta Neurochir Suppl 95:247–251CrossRefGoogle Scholar
  14. 14.
    Czosnyka Z, Czosnyka M, Pickard JD, Chari A (2016) Who needs a revision? 20 years of Cambridge Shunt Lab. Acta Neurochir Suppl 122:347–351. CrossRefPubMedGoogle Scholar
  15. 15.
    Czosnyka Z, Keong N, Kim DJ, Radolovich D, Smielewski P, Lavinio A, Schmidt EA, Momjian S, Owler B, Pickard JD, Czosnyka M (2008) Pulse amplitude of intracranial pressure waveform in hydrocephalus. Acta Neurochir Suppl 102:137–140CrossRefGoogle Scholar
  16. 16.
    Czosnyka ZH, Czosnyka M, Pickard JD (2002) Shunt testing in-vivo: a method based on the data from the UK shunt evaluation laboratory. Acta Neurochir Suppl 81:27–30PubMedGoogle Scholar
  17. 17.
    Dupepe EB, Hopson B, Johnston JM, Rozzelle CJ, Jerry Oakes W, Blount JP, Rocque BG (2016) Rate of shunt revision as a function of age in patients with shunted hydrocephalus due to myelomeningocele. Neurosurg Focus 41:E6. CrossRefPubMedPubMedCentralGoogle Scholar
  18. 18.
    Goeser CD, McLeary MS, Young LW (1998) Diagnostic imaging of ventriculoperitoneal shunt malfunctions and complications. Radiographics 18:635–651. CrossRefPubMedGoogle Scholar
  19. 19.
    Jorgensen J, Williams C, Sarang-Sieminski A (2016) Hydrocephalus and ventriculoperitoneal shunts: modes of failure and opportunities for improvement. Crit Rev Biomed Eng 44:91–97. CrossRefPubMedGoogle Scholar
  20. 20.
    Kanangi SMR, Balasubramaniam C (2018) Shunt infections: a review and analysis of a personal series. Childs Nerv Syst 34:1915–1924. CrossRefPubMedGoogle Scholar
  21. 21.
    Kasprowicz M, Lalou DA, Czosnyka M, Garnett M, Czosnyka Z (2016) Intracranial pressure, its components and cerebrospinal fluid pressure-volume compensation. Acta Neurol Scand 134:168–180. CrossRefPubMedGoogle Scholar
  22. 22.
    Khan AA, Jabbar A, Banerjee A, Hinchley G (2007) Cerebrospinal shunt malfunction: recognition and emergency management. Br J Hosp Med (Lond) 68:651–655CrossRefGoogle Scholar
  23. 23.
    Lalou AD, Czosnyka M, Nabbanja E, Petrella G, Pickard JD, Hutchinson PJ, Czosnyka Z (2018) Outcome and financial implications of shunt infusion studiesGoogle Scholar
  24. 24.
    Lavinio A, Czosnyka Z, Czosnyka M (2008) Cerebrospinal fluid dynamics: disturbances and diagnostics. Eur J Anaesthesiol Suppl 42:137–141. CrossRefPubMedGoogle Scholar
  25. 25.
    Limbrick DD, Jr., Baird LC, Klimo P, Jr., Riva-Cambrin J, Flannery AM, Pediatric Hydrocephalus Systematic R, Evidence-Based Guidelines Task F (2014) Pediatric hydrocephalus: systematic literature review and evidence-based guidelines. Part 4: Cerebrospinal fluid shunt or endoscopic third ventriculostomy for the treatment of hydrocephalus in children. J Neurosurg Pediatr 14 Suppl 1:30–34., 2014PubMedGoogle Scholar
  26. 26.
    Madsen JR, Abazi GS, Fleming L, Proctor M, Grondin R, Magge S, Casey P, Anor T (2011) Evaluation of the ShuntCheck noninvasive thermal technique for shunt flow detection in hydrocephalic patients. Neurosurgery 68:198–205; discussion 205. CrossRefPubMedGoogle Scholar
  27. 27.
    Marmarou A, Shulman K, Rosende RM (1978) A nonlinear analysis of the cerebrospinal fluid system and intracranial pressure dynamics. J Neurosurg 48:332–344. CrossRefPubMedGoogle Scholar
  28. 28.
    Nabbanja E, Pickard JD, Lalou AD, Czosnyka ZH (2018) Use of CSF infusion studies to unblock occluded hydrocephalus ventricular shunt catheters: a preliminary report of two patients. BMJ Case Rep 2018.
  29. 29.
    Paulsen AH, Lundar T, Lindegaard KF (2015) Pediatric hydrocephalus: 40-year outcomes in 128 hydrocephalic patients treated with shunts during childhood. Assessment of surgical outcome, work participation, and health-related quality of life. J Neurosurg Pediatr 16:633–641. CrossRefPubMedGoogle Scholar
  30. 30.
    Petrella G, Czosnyka M, Keong N, Pickard JD, Czosnyka Z (2008) How does CSF dynamics change after shunting? Acta Neurol Scand 118:182–188. CrossRefPubMedGoogle Scholar
  31. 31.
    Petrella G, Czosnyka M, Smielewski P, Allin D, Guazzo EP, Pickard JD, Czosnyka ZH (2009) In vivo assessment of hydrocephalus shunt. Acta Neurol Scand 120:317–323. CrossRefPubMedGoogle Scholar
  32. 32.
    Richards HK, Seeley HM, Pickard JD (2009) Who should perform shunt surgery? Data from the United Kingdom Shunt Registry. Cerebrospinal Fluid Res 6.
  33. 33.
    Robin X, Turck N, Hainard A, Tiberti N, Lisacek F, Sanchez JC, Muller M (2011) pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 12:77. CrossRefPubMedPubMedCentralGoogle Scholar
  34. 34.
    Rudd TG, Shurtleff DB, Loeser JD, Nelp WB (1973) Radionuclide assessment of cerebrospinal fluid shunt function in children. J Nucl Med 14:683–686PubMedGoogle Scholar
  35. 35.
    Sakka L, Chomicki A, Gabrillargues J, Khalil T, Chazal J, Avan P (2016) Validation of a noninvasive test routinely used in otology for the diagnosis of cerebrospinal fluid shunt malfunction in patients with normal pressure hydrocephalus. J Neurosurg 124:342–349. CrossRefPubMedGoogle Scholar
  36. 36.
    Simon TD, Butler J, Whitlock KB, Browd SR, Holubkov R, Kestle JR, Kulkarni AV, Langley M, Limbrick DD Jr, Mayer-Hamblett N, Tamber M, Wellons JC 3rd, Whitehead WE, Riva-Cambrin J, Hydrocephalus Clinical Research N (2014) Risk factors for first cerebrospinal fluid shunt infection: findings from a multi-center prospective cohort study. J Pediatr 164:1462–1468 e1462. CrossRefPubMedPubMedCentralGoogle Scholar
  37. 37.
    Smalley ZS, Venable GT, Einhaus S, Klimo P, Jr. (2017) Low-pressure hydrocephalus in children: a case series and review of the literature. Neurosurgery 80:439–447. CrossRefGoogle Scholar
  38. 38.
    Smielewski P, Czosnyka Z, Kasprowicz M, Pickard JD, Czosnyka M (2012) ICM+: a versatile software for assessment of CSF dynamics. Acta Neurochir Suppl 114:75–79. CrossRefPubMedGoogle Scholar
  39. 39.
    Smielewski P, Lavinio A, Timofeev I, Radolovich D, Perkes I, Pickard JD, Czosnyka M (2008) ICM+, a flexible platform for investigations of cerebrospinal dynamics in clinical practice. Acta Neurochir Suppl 102:145–151CrossRefGoogle Scholar
  40. 40.
    Sorar M, Er U, Ozisik P, Ozeren E, Simsek S (2014) The impact of antibiotic-impregnated catheters on ventriculoperitoneal shunt infection. Turk J Med Sci 44:393–396CrossRefGoogle Scholar
  41. 41.
    Tamber MS, Klimo P, Jr., Mazzola CA, Flannery AM, Pediatric Hydrocephalus Systematic R, Evidence-Based Guidelines Task F (2014) Pediatric hydrocephalus: systematic literature review and evidence-based guidelines. Part 8: Management of cerebrospinal fluid shunt infection. J Neurosurg Pediatr 14 Suppl 1:60–71. doi:, 2014PubMedGoogle Scholar
  42. 42.
    Weerakkody RA, Czosnyka M, Schuhmann MU, Schmidt E, Keong N, Santarius T, Pickard JD, Czosnyka Z (2011) Clinical assessment of cerebrospinal fluid dynamics in hydrocephalus. Guide to interpretation based on observational study. Acta Neurol Scand 124:85–98. CrossRefPubMedGoogle Scholar
  43. 43.
    Widder DJ, Davis KR, Taveras JM (1986) Assessment of ventricular shunt patency by sonography: a new noninvasive test. AJR Am J Roentgenol 147:353–356. CrossRefPubMedGoogle Scholar

Copyright information

© Springer-Verlag GmbH Germany, part of Springer Nature 2019

Authors and Affiliations

  • Sandra Fernandes Dias
    • 1
    • 2
  • Afroditi–Despina Lalou
    • 3
  • Regine Spang
    • 1
  • Karin Haas-Lude
    • 4
  • Matthew Garnett
    • 3
  • Helen Fernandez
    • 3
  • Marek Czosnyka
    • 3
    • 5
  • Martin U. Schuhmann
    • 1
    Email author
  • Zofia Czosnyka
    • 3
  1. 1.Division of Pediatric Neurosurgery, Department of NeurosurgeryUniversity Hospital of TübingenTübingenGermany
  2. 2.Department of NeurosurgeryUniversity Hospital of ZurichZürichSwitzerland
  3. 3.Academic Neurosurgery UnitUniversity of Cambridge School of Clinical Medicine, Addenbrooke’s HospitalCambridgeUK
  4. 4.Department of Pediatric NeurologyUniversity Children’s Hospital TübingenTübingenGermany
  5. 5.Institute of Electronic SystemsWarsaw University of TechnologyWarsawPoland

Personalised recommendations