Even with increasing nonshunt options for

Transcription

1 RESEARCH HUMAN CLINICAL STUDIES Noninvasive Thermal Evaluation of Ventriculoperitoneal Shunt Patency and Cerebrospinal Fluid Flow Using a Flow Enhancing Device Mustafa Q. Hameed, MD David Zurakowski, PhD Mark R. Proctor, MD Scellig S.D. Stone, MD, PhD Benjamin C. Warf, MD Edward R. Smith, MD Liliana C. Goumnerova, MD Marek Swoboda, PhD Tomer Anor, PhD Joseph R. Madsen, MD Department of Neurosurgery, Boston Children s Hospital, Harvard Medical School, Boston, Massachusetts; Department of Anesthesiology, Boston Children s Hospital, Harvard Medical School, Boston, Massachusetts; Drexel University, School of Biomedical Engineering, Science and Health Systems, Philadelphia, Pennsylvania Previous Presentation: 45th Annual Meeting of the AANS/CNS Section on Pediatric Neurological Surgery, December 5-8, 2016, Orlando, Florida. Correspondence: Joseph R. Madsen, MD, Department of Neurosurgery, Boston Children s Hospital, 300 Longwood Avenue, Boston, MA joseph.madsen@childrens.harvard.edu Received, July 10, Accepted, May 29, Copyright C 2018 by the Congress of Neurological Surgeons BACKGROUND: While a noninvasive flow determination would be desirable in the diagnosis of cerebrospinal fluid shunt malfunction, existing studies have not yet defined a role for thermal flow detection. OBJECTIVE: To evaluate a revised test protocol using a micropumper designed to transiently enhance flow during thermal testing to determine whether thermal detection of flow is associated with progression to shunt revision surgery. METHODS: Eighty-two unique tests were performed in 71 shunts. The primary outcome, need for revision within 7 d of testing, was compared with results of micropumperaugmented thermal flow detection. Statistical analysis was based on blind interpretation of test results and raw temperature data recorded during testing. RESULTS: The test was sensitive (73%) and specific (68%) in predicting need for revision, with 5.6-fold higher probability of revision when flow was not detected. Negative predictive value in our sample was 94.2%. The probability of not requiring revision increased with increasing total temperature drop. Analysis of various possible thresholds showed that the optimal temperature cutoff may be lower than suggested by the manufacturer (0.125 Cvs0.2 C). CONCLUSION: This is the first study to report a strong association between thermal flow evaluation and a clinical impression that a shunt is not malfunctioning. The current recommended threshold may increase the false positive rate unnecessarily, and as clinicians gain experience with the method, they may find value in examining the temperature curves themselves. Multicenter studies are suggested to further define a role for this diagnostic test. KEY WORDS: Cerebrospinal fluid, Hydrocephalus, Ventriculoperitoneal shunt, Ventriculoperitoneal shunt flow, Ventriculoperitoneal shunt malfunction Neurosurgery 0:1 10, 2018 DOI: /neuros/nyy246 Even with increasing nonshunt options for hydrocephalus patients, 1-4 the insertion of ventriculoperitoneal (VP) shunts and, consequently, the diagnosis and repair of shunt failure remain among the most common neurosurgical problems in the United States Shunt ABBREVIATIONS: AUC, area under the curve; CSF, cerebrospinal fluid; FD, flow detected; FNC, flow not confirmed; NPV, negative predictive value; OR, odds ratio; ROC, receiver operating characteristic; SSM, suspected shunt malfunction; VP, ventriculoperitoneal malfunction is a critical cause of morbidity and mortality in patients with cerebrospinal fluid (CSF) shunts, and little is known about normal physiological shunt flow The diagnosis of shunt malfunction is difficult to define by prospective diagnostic criteria. Standard methods for investigating shunt status rely either on interpretations of static imaging of the brain, or invasive procedures such as shunt taps, radionuclide studies, and lumbar punctures However, it is well documented that true malfunctions occur without change in ventricular size on imaging, and asymptomatic ventriculomegaly can occur in the NEUROSURGERY VOLUME 0 NUMBER

2 HAMEED ET AL absence of shunt malfunction and require no intervention. 31,32 Not unexpectedly, two different experienced neurosurgeons may reach different conclusions about the need for immediate surgery in any specific instance. This imprecision in the definition of shunt malfunction makes evaluation of diagnostic aids difficult, and may have contributed to the failure of well-designed clinical trials to show advantages of one shunt valve over another. 7,8,33,34 More quantitative, sensitive, and precise diagnostic noninvasive methods would be helpful. The most robust finding at surgery that retrospectively confirms shunt malfunction is an occluded shunt, which is made patent by the surgery, so a preoperative noninvasive test sensitive to flow or patency would be of value. Attempts at noninvasive CSF flow measurement, however, have lacked the reproducibility and convenience needed for clinical settings and require further research and refinement for practical application and clinical use Thermal techniques based on transcutaneous thermal convection are among the most studied and promising methods to evaluate CSF flow in shunts ShuntCheck III R (NeuroDx Development, Yardley, Pennsylvania) is an FDA-approved testing system using a thermal technique and a handheld micropumper to detect and potentially quantify shunt CSF flow. 46 An earlier version of the device and the transcutaneous thermal technique were previously evaluated in patients having routine shunt follow-up as well as in patients having suspected shunt malfunction (SSM), and while the technique was sensitive and specific for detecting flow, the presence of flow did not accurately predict the clinical diagnosis of shunt failure. 17 We now assess the micropumper device, designed to transiently enhance shunt flow during the thermal test, to determine whether this addition to noninvasive thermal flow detection improves the ability to predict the need for shunt revision surgery. METHODS Study Design, Patient Selection, and Clinical Endpoint This prospective research study was done under a research protocol approved by the Institutional Review Board. Informed consent was obtained from a convenience series of patients between the ages of 3 and 20 years with VP shunts, presenting to the neurosurgery outpatient clinic and the emergency room at an academic pediatric tertiary care center before testing was performed (82 unique tests in 71 shunts). Patient presentations consisted both of asymptomatic routine follow-up assessments, as well as symptomatic cases with SSM. Only patients with a distal catheter that was unambiguously palpable as it passed over the clavicle were enrolled in the study. Repeat tests (totaling 11 tests) were included in analyses if time between tests was greater than 1 mo, or if a shunt revision occurred between the tests. The 1-mo interval was considered sufficient, based on clinical experience as well as published neurosurgical papers, 47,48 to allow development or progression of symptoms in cases of SSM. In keeping with prior neurosurgical literature, 47,48 the primary outcome for subjects enrolled in this study was progression to shunt revision surgery based on the attending neurosurgeon s clinical judgment, considering each patient s individual history and current presentation, and the results of imaging and diagnostic studies (other than thermal flow detection) within 7 d. Patients who went on to surgery for reasons other than suspected occlusion were kept in the analysis, and their effect on the results will be noted. Thermal Flow Detection Micropumper The micropumper (Figure 1A) is a hand-held, noninvasive CSF flow generating device made by the manufacturer of the ShuntCheck III R system (NeuroDx Development) that interrogates VP shunts by delivering specific vibration pulses to the dome of the shunt valve. 49 The device consists of a miniature electric motor with an eccentric counterweight, mounted on a plastic foot, and vibrates at 60 Hz for 60 s. The frequency, duration, and duty cycle of the vibration are controlled by an electronic circuit. The micropumper generates controlled, reproducible CSF flow through patent, but not occluded or partially occluded shunts. This temporary increase in flow can then be detected via thermal dilution, as in the test procedure below. The device has received FDA 510(k) approval. 46 Bench-top evaluation by the manufacturer has demonstrated that the use of the micropumper generates robust flow through commonly used commercial shunt valves while keeping the suction pressures generated at the ventricular tip 41% to 82% lower than those generated by manual shunt pumping, and does not adversely affect the valves themselves. Test Procedure All micropumper-augmented thermal flow detection tests were performed with patients in a sitting position. The distal shunt catheter was identified and marked as it passed subcutaneously over the clavicle. An adhesive skin patch with 3 temperature sensors was placed over the clavicle, with the middle (test) sensor directly over the shunt catheter and the 2 remaining (control) sensors positioned symmetrically on either side (Figure 1B). 17 The shunt valve was located and marked for easy access during the latter part of the test. The sensors were then connected to a handheld device that directs the timing and application of subsequent test steps and records continuous temperature data from all 3 sensors, and the test was started. After 10 s of baseline temperature recording, an instant gel ice pack was activated and placed on the skin over the shunt immediately above the adhesive patch for 60 s before being removed. When there is natural flow in the shunt, CSF passing underneath the adhesive patch after being cooled by the ice pack results in a measurably lower temperature reading at the test sensor compared to the control sensors (Figure 1C). On the other hand, absence of flow does not cause any significant difference in temperatures detected by the 3 sensors. One hundred twenty seconds after removal of the ice pack, the same ice pack was reapplied for an additional 60 s. Immediately after removal of the second ice pack application, the activated micropumper was gently placed over the shunt valve and held in place for 60 s (Figure 1D). Temperature data collection continued for 240 s after removal of the massager. Total test duration was 550 s. Thermal data were blindly interpreted by an algorithm that computes the difference in temperature between the test sensor and the mean of the control sensors for each time point, generating a curve of the locally adjusted temperature T(t),wheret is time. 2 VOLUME0 NUMBER

3 NONINVASIVE THERMAL EVALUATION OF SHUNT FLOW FIGURE 1. Placement of micropumper, ice pack and sensors for thermal detection of flow. A, Micropumper. B, Three temperature sensors are placed on the skin over the clavicle with middle sensor over the shunt catheter, and an activated instant gel ice pack is applied above the adhesive patch. C, Area of cooled skin after ice pack removal is shown (shaded area). When there is flow in the shunt, cooled CSF results in a measurable temperature drop as it passes underneath the test sensor. D, The activated micropumper is gently placed over the shunt valve and held in place for 60 s. To understand the relative contributions of the 2 phases of the test, we identified the maximal temperature drop ( T) prior to micropumper application (designated T Natural ) and the additional maximal temperature drop after micropumper application ( T Induced ) and their sum ( T Total ; Figure 2A). Collected data are classified into categorical readings of flow detected (FD), which for this analysis we take as a testable prediction of no malfunction, or flow not confirmed (FNC; Figure 2B), taken as a testable prediction of malfunction. The manufactured devices include a preprogrammed T Total threshold of 0.2 C, based on previous animal and pilot clinical tests. 17,45 We further analyzed the consequences of varying this threshold parameter as will be described below. Statistical Analysis All statistical data analysis was conducted based on the blind interpretation of test results and raw temperature data made available by the manufacturer (NeuroDx Development). Power analysis indicated that the sample size was sufficient to provide 80% power (beta = 0.2) for detecting a median difference of 0.1 C in total temperature drop between groups using the Mann Whitney U-test with a 2-tailed alpha level of 0.05 (nquery Advisor version 7.0, Statistical Solutions, Cork, Ireland). Using a cutoff value for total temperature drop of <0.2 C, we calculated sensitivity and specificity and used logistic regression and Bayes theorem to estimate the odds of shunt revision and the negative predictive value (NPV) of thermal testing, respectively. 50 Comparison of mean temperature drop in subjects that underwent shunt revision surgery and those who did not was performed using the nonparametric Mann Whitney U-test and presented graphically using box-and-whisker plots, where the box is divided at the median and the length of the box represented the interquartile range (25th-75th percentile). The relationship between temperature drop and the probability of shunt revision was determined using maximum likelihood estimation in binary logistic regression with the likelihood ratio test to assess significance. 51 Receiver operating characteristic (ROC) curve analysis was conducted to evaluate the diagnostic utility of the test with area under the curve (AUC) as a measure of accuracy. 52 Two-tailed values of P.05 were considered NEUROSURGERY VOLUME 0 NUMBER

4 HAMEED ET AL FIGURE 3. Flow of study participants. Flow of recruited participants through study protocol. FNC, flow not confirmed; FD, flow detected. TABLE 1. Thermal Prediction of Shunt Malfunction and Progression to Shunt Revision Surgery Shunt revision Yes No Total FIGURE 2. Temperature curves. A, Schematic representation of a thermal curve generated during a micropumper-augmented thermal flow detection test, showing T Natural, T Induced,and T Total parameters. The light gray bars represent first and second ice application, and the dark gray bar indicates micropumper placement. B, Representative thermal curves from 3 patients, showing no natural or induced flow (FNC), no natural flow but >0.2 C induced flow (FD 1), and >0.2 C natural flow as well as >0.2 C induced flow (FD 2). T, temperature drop; FD, flow detected; FNC, flow not confirmed. statistically significant. All analyses were performed using IBM SPSS Statistics v23.0 software (IBM Corporation, Armonk, New York). RESULTS Study Participants Figure 3 shows the flow of recruited participants through study protocol. Thermal flow evaluation was completed in all enrolled patients, with no adverse events. Fifty-four unique tests (65.9%) were performed on asymptomatic patients presenting for routine follow-up assessment, and 28 tests (34.1%) were conducted in cases with SSM. In SSM cases, 11 tests (39.3%) progressed to shunt revision surgery. Eight patients were tested more than once (19 unique tests). Of these, 7 patients received a single repeat test during regular postoperative evaluation following shunt revision Predicted Malfunction + (FNC) (FD) Total FD, flow detected; FNC, flow not confirmed Outcome comparison shows a sensitivity of 73% and a specificity of 68%. The odds of shunt revision surgery following a predicted malfunction (FNC) on thermal testing were 5.57 times higher compared to a FD result (OR: 5.57, P =.01). Probability of patients with FD not progressing to shunt revision surgery (NPV) was 94.1% in our sample, where prevalence of shunt revision surgery was 13.41%. surgery. In addition, 4 tests were performed on a single patient, each spaced at least 1 mo apart while a fifth test was performed on this patient during a regular postrevision visit. Only 2 patients crossed over from FNC to FD on repeat testing; none changed from FD to FNC. Thermal Flow Detection With Micropumper Thermally Detectable Flow and Progression to Surgery The observed prevalence of shunt revision surgery in our sample was 13.41%. Outcome (+malfunction/fnc or malfunction/fd; +/ shunt revision surgery) comparison in a 2 2 contingency analysis indicated a sensitivity of 73% and a specificity of 68% when the manufacturer recommended test threshold was used (Table 1). The odds of shunt revision surgery following a +malfunction/fnc test result were 5.6 times higher compared to a malfunction/fd result (odds ratio [OR]: 4 VOLUME0 NUMBER

5 NONINVASIVE THERMAL EVALUATION OF SHUNT FLOW TABLE 2. Estimated Positive and Negative Predictive Values of Thermal Flow Detection With Variations in Population Prevalence of Actual Revision (A) T Total 0.2 C (B) T Total C Prevalence (%) PPV (%) NPV (%) PPV (%) NPV (%) PPV, positive predictive value; NPV, negative predictive value; T Total, total temperature drop Estimated positive and Negative Predictive Values using Bayes theorem for a range of possible shunt revision prevalence, using T Total threshold of (A) <0.2 C and (B) <0.125 C. Values for the current study sample are in bold, and italicized. 5.57, P =.01). In addition, the probability of malfunction/fd not progressing to shunt revision surgery (NPV) was 94.2% in our study sample. 50 Since negative and positive predictive values actually depend on the true prevalence of the condition, these values were also estimated for different hypothetical patient populations with varying prevalence of shunt revision surgeries (Table 2, columna). Magnitude of Temperature Drop ( T) is Significantly Lower in Patients who Require Shunt Revision Surgery The median values of all 3 thermally detected flow parameters, T Total (0.07vs0.29,P =.004), T Natural (0.05vs0.17, P =.05), and T Induced (0.00 vs 0.11, P =.018) were significantly lower in the shunts that subsequently underwent revision compared to those that did not (Figures 4A-4C). The Probability of not Requiring Shunt Revision Surgery Increases with Oncreasing T Total Binary logistic regression analysis using the likelihood ratio test indicates a significant relationship between greater temperature drop (reflecting more flow) and the probability of being discharged without a shunt revision (likelihood ratio test = 8.84, P =.003, Figure 5). Thermal Detection of VP Shunt Flow has Diagnostic Utility as a Means of Predicting that no Acute Intervention is Needed An ROC curve visually illustrates the performance of a dichotomous test as the criterion is varied, and is created by plotting the test s sensitivity (true positive rate) against 1-specificity (false positive rate) at various threshold settings. A chance diagonal represents a useless test, and points above this line indicate better than random results with the curve of a perfect test running along the left and upper borders of the plot area. 52 Analysis of the curve for micropumper-augmented thermal flow detection shows that T Total is a fair predictor of whether shunt revision surgery is needed (AUC: 0.773, 95% confidence interval: , P =.004). Youden s J statistic (J t = Sensitivity t + Specificity t 1) for each threshold t was calculated, and the maximum J t identified the cutoff ( T Total : C) at which the ROC curve is furthest away from the chance diagonal, representing the optimal balance between sensitivity (73%) and specificity (80.28%; Figure 6). Using this cutoff increases the NPV of the test in our sample to 95.1%, and similarly across the range of estimated prevalence of shunt revision (Table 2, columnb). DISCUSSION We report for the first time the diagnostic utility of thermally detected VP shunt flow in the management of chronically shunted patients presenting to a pediatric neurosurgical department, using a cross-sectional observational study design. One of the most robust and compelling findings in this study was the prediction of nonprogression to surgery by thermal flow detection with micropumper. The test correctly identified no flow (+malfunction/fnc) in 8 out of 11 cases where the patient went on to shunt revision surgery (true positive, sensitivity 73%), and flow ( malfunction/fd) in 48 out of 71 shunts that did not subsequently require revision (true negative, specificity 63%). The odds of shunt revision surgery following FNC tests were 5.57 times higher compared to FD tests. Significantly lower natural, induced, and total temperature drops were seen in patients who progressed to surgery compared to those who did not (Figures 4A-4C). On the other hand, the NPV, which is the probability of not progressing to shunt revision surgery after a negative test ( malfunction/fd) result, was 94.2%. Additionally, the probability of not requiring revision surgery increased with increasing temperature drops (reflecting more flow) in our study population (Figure 5). In our opinion, the true utility of the test may lie in correctly identifying patients presenting with SSM who do not require immediate intervention a rule-out test. There were 3 cases where flow was identified by the device that nevertheless progressed to shunt revision surgery. However, none of these patients went to the operating room to replace obstructed shunts one was a case of symptomatic over-drainage, another required valve replacement following mechanical failure of an adjustable shunt, and the third required surgery to repair a detached Rickham reservoir with subsequent subcutaneous collection of CSF. Post hoc exclusion of these patients would decrease the false negative rate to zero and increase both the sensitivity and NPV of the test to 100%. We had not stipulated exclusion of such patients in the study design, so they are included in our analyses, effectively making the statistical NEUROSURGERY VOLUME 0 NUMBER

6 HAMEED ET AL FIGURE 4. Thermal flow parameters in surgical vs nonsurgical patients. Median temperature drops: A, T Total (0.07 vs 0.29, P =.004); B, T Natural (0.05 vs 0.17, P =.05); and C, T Induced (0.00 vs 0.11, P =.018), were significantly lower in patients who subsequently underwent shunt revision surgery compared to those who did not. Boxes indicate median and 25th and 75th percentiles. Whiskers represent 2.5th and 97.5th percentiles. Outlier values are indicated by solid circles; T Total, total temperature drop; T Natural, natural temperature drop; T Induced, induced temperature drop. bar more rigorous for this prospective study. Subsequent study designs, which exclude such patients, could yield higher sensitivity and NPVs. However, this does point to a qualification on the interpretation of the test: robust flow itself may be an indication for surgery in some cases of over-drainage, or persistent flow through a tenuous fibrous tract after shunt fracture. We believe it is likely that thermal flow detection will best be used in conjunction with clinical impressions or other tests, and not as a definition of shunt malfunction in isolation. In our cohort, 23 shunts with FNC did not progress to surgery. There may be several plausible explanations for the seemingly high incidence of the finding of FNC in the nonsurgical group. For one, noise resulting from patient movement could obscure a flow signal. In addition, it is possible for flow to be below threshold for detection by the device, which is, of course, true for all measuring instruments. For principal binary outcome event comparison, we used the manufacturer recommended temperature threshold in classifying +malfunction/fnc vs malfunction/fd results. This cutoff, as used in previous clinical studies, 17,53 had been conservatively chosen to minimize the number of false negatives and avoid missing patients with potentially obstructed shunts. Indeed, the ROC curve analysis of our sample indicates that while thermal flow detection has diagnostic utility as a dichotomous test using the manufacturer recommended threshold (0.2 C), the algorithm could use a lower threshold (0.125 C) to increase specificity to 80% without sacrificing sensitivity (73%). Subsequent randomized multicenter trials with larger and more representative sample sizes could potentially identify even smaller thresholds that may further increase both the specificity and sensitivity of the test. Alternatively, perhaps some of the patients tested were truly shunt independent, with VP shunts that no longer flowed. We have previously demonstrated the intermittency of shunt flow consistent with data in the literature suggesting the intermittent nature of CSF flow These explanations are not mutually exclusive and may be resolved with serial retesting of chronically shunted patients. Our data suggest that thermal flow measurement, along with a means to transiently and safely induce flow, may be useful as an additional, useful clinical data point when following chronically shunted patients in the clinic. Since clinical decision-making relies on the context of any given decision, neurosurgeons may also 6 VOLUME0 NUMBER

7 NONINVASIVE THERMAL EVALUATION OF SHUNT FLOW within a patient may be more predictive of clinical trajectory than a single isolated value. Valuable insight into why some hydrocephalic patients need multiple shunt revisions while others may not require surgery for decades might be gained from serially testing shunt flow in individual patients using this relatively inexpensive and noninvasive technique, in future studies. FIGURE 5. Thermal detection of flow and probability of no shunt revision. The probability of not progressing to shunt revision surgery increases with increasing T Total detected by the thermal test. Dotted lines represent 95% confidence interval (likelihood ratio test = 8.84, P =.003). T Total, total temperature drop. Study Limitations Clinicians could access the raw temperature curves generated by the test software if desired (although they did so infrequently) as allowed by the Institutional Review Board. However, they were instructed to disregard the thermal data during decision making, and all reported doing so. To avoid subconscious bias in surgical decisions, strict operator blinding could be achieved with specially designed (blinded) testing hardware, but such was not done in this study. The major modification of this study procedure over that published earlier was the addition of a second thermal application with a flow-enhancing micropumper. 17 While the sum of the baseline and micropumper enhanced flow succeed in clarifying the probability that a patient would go to surgery, it is not clear from these data whether this results from the micropumper activity per se, or the doubled time of cold exposure to increase the signal to noise ratio. Since the curves (Figure 2B) showed cases where there was no temperature drop with the first cold pack application, but presented temperature drops when the micropumper was added, there is likely to be an effect of both changes in the protocol. While the current study implies a role for thermal detection of flow in the management of chronically shunted patients presenting to a pediatric neurosurgical department, rigorous clinical evaluation in a multicenter, prospective, double-blinded clinical trial with a large sample size and better stratification of pretest clinical status is required to provide a sharper estimation of the utility offered by this thermal approach to decision making in addition to clinical and imaging decision. CONCLUSION FIGURE 6. ROC curve for micropumper-augmented thermal flow detection. ROC curve analysis shows that T Total is a fair predictor of need for shunt revision surgery (AUC: 0.773, 95% confidence interval: , P =.004). on curve corresponds to current threshold of 0.2 C, while represents the threshold, which results in the highest test sensitivity (73%) and specificity (80.28%), and corresponds to a T Total cutoff of C. ROC, receiver operating characteristic; AUC, area under the curve; CI, confidence interval; T Total, total temperature drop. find value in viewing and comparing graphical temperature data from serial thermal testing in individual patients to form conclusions, rather than simply rely on the tests binary results. The skewness in the distribution of temperature measurements may indicate the possibility that trends or changes in flow dynamics While prior studies suggested that flow in shunts can be detected by a thermal technique, this is the first study, utilizing 2 thermal exposures and a flow-enhancing micropumper, to show statistically significant results using the binary FD or FNC outputs of the device. The current temperature cutoff of 0.2 C, while intended to minimize false negatives, may increase the false positive rate unnecessarily, and experienced clinicians may also become increasingly skillful at interpreting the temperature curves themselves. It is likely that thermal flow detection will find one of its key values in the confirmation of clinical impressions or other tests suggesting that a shunt is not malfunctioning. Such questions of clinical utility should be tested in future studies. Further planned analyses include results of operator blinded studies and influences of a priori clinical impressions on the utility of the result. NEUROSURGERY VOLUME 0 NUMBER

8 HAMEED ET AL Disclosures Funding was received from NIH Phase I SBIR # 5R43HD (PI: Marek Swoboda, PhD/NeuroDx Development, LLC; subcontract to Joseph R. Madsen, MD/Boston Children s Hospital). Dr Madsen and Dr Swoboda are coinventors of the micropumper device (US patent # ), with intellectual property assigned to Boston Children s Hospital and NeuroDx Development, respectively. REFERENCES 1. Demerdash A, Rocque BG, Johnston J, et al. Endoscopic third ventriculostomy: a historical review. Br J Neurosurg. 2017;31(1): Weil AG, Westwick H, Wang S, et al. Efficacy and safety of endoscopic third ventriculostomy and choroid plexus cauterization for infantile hydrocephalus: a systematic review and meta-analysis. Childs Nerv Syst. 2016;32(11): Stone SS, Warf BC. Combined endoscopic third ventriculostomy and choroid plexus cauterization as primary treatment for infant hydrocephalus: a prospective North American series. J Neurosurg Pediatr. 2014;14(5): Kulkarni AV, Riva-Cambrin J, Holubkov R, et al. Endoscopic third ventriculostomy in children: prospective, multicenter results from the Hydrocephalus Clinical Research Network. J Neurosurg Pediatr. 2016;18(4): Patwardhan RV, Nanda A. Implanted ventricular shunts in the United States: the billion-dollar-a-year cost of hydrocephalus treatment. Neurosurgery. 2005;56(1): ; discussion Simon TD, Riva-Cambrin J, Srivastava R, et al. Hospital care for children with hydrocephalus in the United States: utilization, charges, comorbidities, and deaths. J Neurosurg Pediatr. 2008;1(2): Drake JM, Kestle JR, Milner R, et al. Randomized trial of cerebrospinal fluid shunt valve design in pediatric hydrocephalus. Neurosurgery. 1998;43(2): ; discussion Kestle J, Drake J, Milner R, et al. Long-term follow-up data from the Shunt Design Trial. Pediatr Neurosurg. 2000;33(5): Garton HJ, Kestle JR, Drake JM. Predicting shunt failure on the basis of clinical symptoms and signs in children. JNeurosurg. 2001;94(2): Symss NP, Oi S. Is there an ideal shunt? A panoramic view of 110 years in CSF diversions and shunt systems used for the treatment of hydrocephalus: from historical events to current trends. Childs Nerv Syst. 2015;31(2): Wong JM, Ziewacz JE, Ho AL, et al. Patterns in neurosurgical adverse events: cerebrospinal fluid shunt surgery. Neurosurg Focus. 2012;33(5):E Browd SR, Gottfried ON, Ragel BT, Kestle JR. Failure of cerebrospinal fluid shunts: part II: overdrainage, loculation, and abdominal complications. Pediatr Neurol. 2006;34(3): Browd SR, Ragel BT, Gottfried ON, Kestle JR. Failure of cerebrospinal fluid shunts: part I: obstruction and mechanical failure. Pediatr Neurol. 2006;34(2): Staal MJ, Meihuizen-de Regt MJ, Hess J. Sudden death in hydrocephalic spina bifida aperta patients. Pediatr Neurosurg. 1987;13(1): Gmeiner M, Wagner H, Zacherl C, et al. Long-term mortality rates in pediatric hydrocephalus-a retrospective single-center study. Childs Nerv Syst. 2017;33(1): Hanak BW, Bonow RH, Harris CA, Browd SR. Cerebrospinal fluid shunting complications in children. Pediatr Neurosurg. 2017;52: Madsen JR, Abazi GS, Fleming L, et al. Evaluation of the ShuntCheck noninvasive thermal technique for shunt flow detection in hydrocephalic patients. Neurosurgery. 2011;68(1): ; discussion Zorc JJ, Krugman SD, Ogborn J, Benson J. Radiographic evaluation for suspected cerebrospinal fluid shunt obstruction. Pediatr Emerg Care. 2002;18(5): Pitetti R. Emergency department evaluation of ventricular shunt malfunction? Pediatr Emerg Care. 2007;23(3): Park DB, Hill JG, Thacker PG, et al. The role of limited head computed tomography in the evaluation of pediatric ventriculoperitoneal shunt malfunction. Pediatr Emerg Care. 2016;32(9): Maller VV, Agarwal A, Kanekar S. Imaging of ventricular shunts. Semin Ultrasound, CT MR. 2016;37(2): Sivaganesan A, Krishnamurthy R, Sahni D, Viswanathan C. Neuroimaging of ventriculoperitoneal shunt complications in children. Pediatr Radiol. 2012;42(9): Sood S, Canady AI, Ham SD. Evaluation of shunt malfunction using shunt site reservoir. Pediatr Neurosurg. 2000;32(4): Howman-Giles R, McLaughlin A, Johnston I, Whittle I. A radionuclide method of evaluating shunt function and CSF circulation in hydrocephalus. JNeurosurg. 1984;61(3): Chervu S, Chervu LR, Vallabhajosyula B, et al. Quantitative evaluation of cerebrospinal fluid shunt flow. JNuclMed. 1984;25(1): Rocque BG, Lapsiwala S, Iskandar BJ. Ventricular shunt tap as a predictor of proximal shunt malfunction in children: a prospective study. J Neurosurg Pediatr. 2008;1(6): Lee JB, Ahn HY, Lee HJ, Yang JH, Yi JS, Lee IW. Cerebrospinal fluid lumbar tapping utilization for suspected ventriculoperitoneal shunt under-drainage malfunctions. J Korean Neurosurg Soc. 2017;60(1): Winston KR, Lopez JA, Freeman J. CSF shunt failure with stable normal ventricular size. Pediatr Neurosurg. 2006;42(3): Iskandar BJ, McLaughlin C, Mapstone TB, Grabb PA, Oakes WJ. Pitfalls in the diagnosis of ventricular shunt dysfunction: radiology reports and ventricular size. Pediatrics. 1998;101(6): Barnes NP, Jones SJ, Hayward RD, Harkness WJ, Thompson D. Ventriculoperitoneal shunt block: what are the best predictive clinical indicators? Arch Dis Child. 2002;87(3): Jernigan SC, Stone SS, Aronson JP, Putman M, Proctor MR. Episodic ventriculomegaly due to hypernatremia mimicking shunt malfunction: case report. J Neurosurg Pediatr. 2015;16(4): Andres RH, Pendharkar AV, Kuhlen D, Mariani L. Ventricular enlargement due to acute hypernatremia in a patient with a ventriculoperitoneal shunt. JNeurosurg. 2010;113(1): Klimo P, Jr, Thompson CJ, Drake J, Kestle JR. Assessing the validity of the endoscopic shunt insertion trial: did surgical experience affect the results? J Neurosurg. 2004;101(2 suppl): Drake JM, Kestle JR, Tuli S. CSF shunts 50 years on - past, present and future. Childs Nerv Syst. 2000;16(10-11): Lam KW, Drake JM, Cobbold RS. Noninvasive cerebrospinal fluid shunt flow measurement by Doppler ultrasound using ultrasonically excited bubbles: a feasibility study. Ultrasound Med Biol. 1999;25(3): Frank E, Buonocore M, Hein L. Magnetic resonance imaging analysis of extremely slow flow in a model shunt system. Childs Nerv Syst. 1992;8(2): Aralar AJ, Bird MD, Graham RD, et al. Ultrasound characterization of interface oscillation as a proxy for ventriculoperitoneal shunt function. Conf Proc IEEE Eng Med Biol Soc Aug;2016: Hartman R, Aglyamov S, Fox DJ, Jr, Emelianov S. Quantitative contrastenhanced ultrasound measurement of cerebrospinal fluid flow for the diagnosis of ventricular shunt malfunction. JNeurosurg. 2015;123(6): Martin AJ, Drake JM, Lemaire C, Henkelman RM. Cerebrospinal fluid shunts: flow measurements with MR imaging. Radiology. 1989;173(1): Stein SC, Apfel S. A noninvasive approach to quantitative measurement of flow through CSF shunts. JNeurosurg. 1981;54(4): Go KG, Lakke JP, Beks JW. A harmless method for the assessment of the patency of ventriculoatrial shunts in hydrocephalus. Dev Med Child Neurol. 1968;10(s16): Go KG, Melchior HJ, Lakke JP. A thermosensitive device for the evaluation of the patency of ventriculo-atrial shunts in hydrocephalus. Acta Neurochir. 1968;19(4): Chiba Y, Yuda K. Thermosensitive determination of CSF shunt patency with a pair of small disc thermistors. JNeurosurg. 1980;52(5): Chiba Y, Ishiwata Y, Suzuki N, Muramoto M, Kunimi Y. Thermosensitive determination of obstructed sites in ventriculoperitoneal shunts. JNeurosurg. 1985;62(3): Neff S. Measurement of flow of cerebrospinal fluid in shunts by transcutaneous thermal convection. Technical note. JNeurosurg. 2005;103(4 suppl): ShuntCheck III 510(k) Premarket notification, Food and Drug Administration. Available at: pdf12/k pdf. Accessed January 15, Yue EL, Meckler GD, Fleischman RJ, et al. Test characteristics of quick brain MRI for shunt evaluation in children: an alternative modality to avoid radiation. J Neurosurg Pediatr. 2015;15(4): Boyle TP, Paldino MJ, Kimia AA, et al. Comparison of rapid cranial MRI to CT for ventricular shunt malfunction. Pediatrics. 2014;134(1):e47-e54. 8 VOLUME0 NUMBER

9 NONINVASIVE THERMAL EVALUATION OF SHUNT FLOW 49. Swoboda M, Hochman MG, Mattiucci ME, Fritz FJ, Madsen JR, inventors; Children s Medical Center Corp, ShuntCheck Inc., assignees. CSF shunt flow enhancer, method for generating CSF flow in shunts and assessment of partial and complete occlusion of CSF shunt systems. US Patent 9,138,568. United States Patent and Trademark Office. September 22, Fletcher RW, Fletcher SW. In: Clinical Epidemiology: The Essentials. 4thed. Philadelphia, PA: Lippincott Williams & Wilkins; 2005: Hosmer DW, Lemeshow S. In: Applied Logistic Regression.2nded.NewYork,NY: John Wiley; 2000: Pepe MS. In: The Statistical Evaluation of Medical Tests for Classification and Prediction. Oxford, UK: Oxford University Press; 2003: Marlin AE, Gaskill SJ. The use of transcutaneous thermal convection analysis to assess shunt function in the pediatric population. Neurosurgery. 2012;70(2 Suppl Operative): Kadowaki C, Hara M, Numoto M, Takeuchi K. Factors affecting cerebrospinal fluidflowinashunt.br J Neurosurg. 1987;1(4): Hidaka M, Matsumae M, Ito K, Tsugane R, Suzuki Y. Dynamic measurement of the flow rate in cerebrospinal fluid shunts in hydrocephalic patients. Eur J Nucl Med. 2001;28(7): Drake JM, Sainte-Rose C, DaSilva M, Hirsch JF. Cerebrospinal fluid flow dynamics in children with external ventricular drains. Neurosurgery. 1991;28(2): Kadowaki C, Hara M, Numoto M, Takeuchi K, Saito I. CSF shunt physics: factors influencing inshunt CSF flow. Child s Nerv Syst. 1995;11(4): COMMENTS This is an interesting paper as it begins to sketch a picture of how shunt controlled hydrocephalus may be managed in the future. This work will not doubt inform strides to be taken by industry and perhaps in the public domain by patients and families. As is happening in the type 1 diabetes world, patients are working with experts to define the technology evolution that we see today and will no doubt mature in the near future; continuous glucose monitoring integrated with insulin pumps that remove the limitations and the burden of the illness. For those fortunate individuals, current technology allows them to be less conscious of their disease as they live more normal lives. If cerebrospinal fluid drainage remains a mainstay of hydrocephalus control, similar principles will likely be brought to bear; individualized monitoring and drainage control. Patients will no doubt play a greater role in the technology evolution and may do this in the public domain with lesser involvement of industry. Indeed, this is where valved shunting began with Holter and Dahl, and Till. This study adds to our understanding of shunt behavior and what flow means and does not. As with diabetes, we will see technology evolve to support individualized management and once automated, hydrocephalus will have less impact on the lives of our patients. Until then, clinical judgement by patients, parents and surgeons supplemented by imaging and technologies as described will be the mainstays of management. D. Douglas Cochrane Vancouver, Canada The assessment of shunt function remains difficult. The most basic approach is to observe for change in ventricular size on a CT or MRI scan in the setting of appropriate symptoms (progressive headache, emesis, sleepiness/lethargy) but this simplistic and potentially dangerous approach misses a significant number of shunt failures. This paper further develops the use of a previously reported commercial product called Shunt-CheckIII to utilize thermal variance to measure flow in a ventricular shunt. The tract of the shunt is identified and a cooling source applied over it. Changes in temperature downstream are detected and correlated with flow. Previously the existence or absence of flow (as reflected by detection of transcutaneous thermal variance) was detectable but did not accurately predict the need for ventricular shunt revision. In this study they augment the previous device with a micro pumping device which transcutaneously stimulates the valve with a pulsation frequency of 40 stimulations per second. Subjects for the study were asymptomatic patients with ventricular shunts attending the Pediatric Neurosurgery clinic and symptomatic patients evaluated in Emergency Room. Absolute measurements of temperature and the degree of temperature change were made. The clinical endpoint was need for shunt revision within 5 days. This is an imperfect, non-objective endpoint but a single surgeon made operative decisions based on standardized criteria. Statistical evaluation occurred with blinded data. The authors found that thermal changes measured by the device predicted a need for shunt surgery with a 73% sensitivity and 68% specificity. The negative predictive value (ie, no need for intervention) was 94%. The magnitude of temperature change was found to correlate with need for shunt revision. Patients demonstrated a 5.6 times greater likelihood of requiring shunt revision when flow was not detected. As such the authors conclude that the device may have clinical utility in determining those patients who do not require intervention. There is value in a clinical tool that corroborates a clinical impression that no intervention needs to be undertaken. In a difficult clinical setting of a shunt patient with conflicting clinical data a test that provides insight of function would be valuable. Most neurosurgeons with extensive experience with ventricular shunts have a low threshold for exploring shunts but significant variability and practice preferences exist between experienced centers. This device may be of value in further corroborating a clinical impression and may help prevent unnecessary interventions. As such the principal potential clinical benefit at present is likely to rule out obstruction rather than to rule it in. There is potential utility of the device in providing more information about basic flow dynamics of shunt systems. Such information is badly needed to gain better insight into the natural history of flow dynamics of ventricular shunts and there is not a better available technique to my knowledge at present. However, some significant challenges remain: 1) The fundamental premise that flow reflects function is challenging in that we simply do not know what degree of flow is necessary within a shunt for a child with a shunt. It is highly likely that the amount of flow through perfectly functional shunts varies considerably. There have never been normative flow curves published for shunts in humans or animal models. I suspect that the rate of flow through most functional shunts is small and symptomatic shunt failure results from a) subtle reductions in this drainage or b) stresses from other body system that cross a critical threshold thus prompting symptoms of shunt failure (such as are known to occur in select patients with other physiologic stresses like urinary tract infections, severe constipation etc). 2) The temperature differences observed are statistically significant but they remain very small absolute differences. As shown in the diagrams the absolute difference between values is 0.12 C. Even if the device works perfectly I can imagine significant confounders in the difficult real clinical domain of providing clinical care to frightened, hyperactive, developmentally delayed children that could seriously confound these results. The addition of the pumping device to augment flow results further challenges feasibility in the real life world of assessing children. 3) I would have valued more detailed information about surgical findings of shunt obstruction at surgery. Observations and correlations NEUROSURGERY VOLUME 0 NUMBER

10 HAMEED ET AL of ventricular catheter obstruction would seem like important variables that are available without exposing the patient to any additional risk. I would strongly encourage the authors to keep detailed observations of the obstructed segment of shunt at the time of surgery for future studies. In summary, this is a clever concept but challenges remain. The history of hydrocephalus is replete with a variety of different devices that were designed to contribute information about shunt function. None have persisted to remain widely used tools in the assessment of shunt function. There is potential in this device to corroborate a clinical impression of no need for intervention and to provide important natural history flow information on ventricular shunts. The authors are to be congratulated for tackling a difficult problem and pursuing a non-invasive tool for detecting shunt malfunction. Jeffrey P. Blount Birmingham, Alabama The effort by this group shows a thoughtful attempt to add to the current methods of detecting shunt malfunction and selecting patients in need of surgical revision. The study design that allowed for a significant separation between the results of their augmented thermal flow test, and the decision for surgery, is a very positive aspect of the report. I also appreciate the inclusion of a group of patients with mechanical shunt problems which had to be revised, because their study design did not take that situation into account. The authors note, that if they had excluded those patients, the sensitivity would be 100%. This transparency in reporting is commendable and increases both the veracity and utility of their results. Ann Marie Flannery Lafayette, Louisiana 10 VOLUME0 NUMBER