Normal pressure hydrocephalus

• Normal pressure hydrocephalus is a form of communicating hydrocephalus in which the cerebral ventricles enlarge, yet the measured pressure of the cerebrospinal fluid stays within the normal range.
• It overwhelmingly affects older adults and declares itself through a triad: a characteristic gait and balance problem, cognitive slowing with a frontal pattern, and urinary urgency or incontinence.
• Brain imaging shows enlarged ventricles, often with a pattern of tight high convexity and widened Sylvian fissures called disproportionately enlarged subarachnoid-space hydrocephalus (DESH).
• The diagnosis is secured, and shunt surgery recommended, only after physiologic testing such as a CSF tap test or external lumbar drainage.
• Shunt surgery helps a large majority of properly selected patients and can improve all three symptoms, including cognition.

By the end of this page you should be able to:

1. Define normal pressure hydrocephalus and distinguish idiopathic from secondary forms.
2. Explain the pathophysiology from impaired cerebrospinal fluid dynamics to the clinical triad.
3. Recognise the typical gait, cognitive and urinary features and the neuroimaging findings, including DESH and the Evans ratio.
4. Choose between a tap test, external lumbar drainage and infusion testing, and interpret their predictive values.
5. Outline guideline-based management, valve choice and the expected short- and long-term outcomes.
6. List the major mimics and the shunt complications that need long-term follow-up.

This page deals with idiopathic normal pressure hydrocephalus (iNPH), the most common hydrocephalus syndrome of adults. The topic matters because it is one of the few genuinely reversible causes of dementia and gait failure in older people, and it is systematically underdiagnosed: in western Sweden only two to three people per 100,000 with the condition underwent shunt surgery each year, while population studies found many more affected ^[3]. The international guidelines for iNPH were published by Relkin and colleagues in 2005, the Japanese guidelines are now in their third edition (2021), and the American Academy of Neurology (AAN) has a practice guideline on shunt response and its predictors ^[1][2][4]. Where these sources differ this page says so.

The original description by Hakim and Adams in 1965 stressed exactly this paradox: ventricles that were too large for the intracranial pressure, producing a clinical picture that looked like much commoner neurodegenerative disease ^[2]. The pressure is not truly "normal" in every patient: B-waves, raised pulse amplitudes and elevated outflow resistance occur, and some patients have pressures above 200 mmH2O. "Normal-pressure hydrocephalus" is therefore a historical and clinical label rather than a precise physiological statement ^[1].

Normal pressure hydrocephalus is divided into two broad groups ^[3]:

• Idiopathic (iNPH) : no identifiable cause. This is the subject of this page.
• Secondary (sNPH) : follows an identifiable insult such as subarachnoid haemorrhage, bacterial meningitis, severe head injury, brain irradiation, or a previous neurosurgical procedure.

There are also bordering categories that the guidelines keep separate: congenital or developmental forms that remained silent into older age (longstanding overt ventriculomegaly in adults, LOVA, and panventriculomegaly with a wide foramen of Magendie), and familial forms. These often respond to shunting too, but they are not iNPH because an earlier cause or structural variant is present ^[1].

The diagnostic stages used before surgery matter, because the final diagnosis is only confirmed by a good response to a shunt ^[1]:

• Possible iNPH : at least one symptom of the triad, insidious onset, ventriculomegaly on imaging, CSF pressure of 200 mmH2O or less with normal CSF content, and no better explanation. ^[1]
• Probable iNPH : possible iNPH plus either DESH on MRI, or improvement after a tap test or drainage test.
• Definite iNPH : probable iNPH plus objective improvement after CSF shunt surgery (a "shunt responder").

Asymptomatic people with ventriculomegaly and DESH on MRI exist and are called AVIM. In a Japanese community study, four of eight such people developed iNPH over four to eight years, and in a hospital-based study 27 of 52 developed iNPH within three years ^[1].

iNPH is commoner than it is diagnosed. In Scandinavia the population prevalence was about 3.7% overall, rising to 5.9% to 8.9% in those aged 80 years and above on two Swedish surveys using head CT ^[1]; a western Swedish study gave 0.2% for ages 70 to 79 years and 5.9% for age 80 years and above ^[3]. A 2018 meta-analysis gave a mean prevalence of 175 per 100,000 worldwide and 400 per 100,000 in those aged 80 years and above ^[1]. The condition appears commoner in Japan: a mean prevalence of 660 per 100,000 there, against 53 per 100,000 in countries outside Japan, although diagnostic practice and case-finding differ ^[1]. One Norwegian study estimated an incidence of 5.5 per 100,000 ^[2].

Demographically, most reviews cite an average age of onset of about 70 years, whereas the Japanese third-edition series reports an average age at onset of about 75 years; men and women are affected in roughly equal numbers ^[3][1]. If the Scandinavian prevalence applies to the United States, roughly 700,000 people there may have iNPH, more than the number usually quoted for multiple sclerosis, yet only a fraction are ever shunted ^[3]. The most likely reason is difficulty separating iNPH from common comorbidities in the same age group.

iNPH is idiopathic, so there is no single cause. The clues to why it appears lie in the risk factors for communicating hydrocephalus that the guidelines ask you to exclude or to weigh ^[4]:

• prior meningitis or encephalitis
• head injury, including concussion
• subarachnoid haemorrhage
• brain irradiation
• previous intracranial surgery
• a congenital component, suggested by an enlarged head circumference

A family history occasionally matters, and vascular risk factors and small-vessel disease frequently coexist. There is no known modifiable behavioural risk factor, and prevention in the lifestyle sense does not exist: the realistic aim is timely recognition of an affected older adult rather than prevention of the disease.

The problem is cerebrospinal fluid dynamics, not overproduction. In most patients the impairment is at the absorption stage, either at the arachnoid granulations or in the periventricular parenchyma, so resistance to outflow rises and the pressure waveform becomes abnormal even though the resting pressure stays near normal ^[1][3]. Ventricles enlarge because the fluid is not cleared, then the periventricular white matter and the long fibre tracts serving gait and continence are stretched and distorted.

This explains the clinical pattern more than any single anatomical lesion. The frontal-subcortical circuits that govern initiation of gait, executive function and bladder control run close to the ventricles and are sensitive to a subcritical ischaemia of the periventricular end arteries, producing dysfunction rather than infarction. That is why the symptoms are reversible with fluid diversion and why vascular comorbidity makes the recovery smaller. Cerebrospinal fluid removal acutely reverses the mechanical stretch and reduces the pressure waveform, which is why a tap test predicts a shunt response ^[3].

Gait trouble is usually the first and worst symptom, and iNPH should be suspected in any older adult with unexplained, symmetric gait impairment ^[3]. The full triad is not required: many published series and guidelines state that nearly all patients have gait impairment, but only about 60% have the complete triad, and one Japanese survey of 1,524 patients found all three symptoms in only 12.1% ^[1][4].

Typical gait features include difficulty rising from a chair, failure to initiate gait, shuffling with poor foot clearance, festination, tripping or falling, unstable multistep turns, and retropulsion or anteropulsion ^[4]. The gait is choreiform and "magnetic", with short, broad-based steps. Unsteadiness on turning is often the most noticeable thing in the clinic.

Cognitive problems have a frontal-subcortical flavour: slowed processing, impaired executive function, difficulty with planning and problem solving, apathy and amotivation, daytime somnolence and psychomotor slowing ^[4]. Memory retrieval is impaired but recognition memory is relatively preserved, which helps separate iNPH from Alzheimer disease. Language is spared in the sense that naming and comprehension remain intact early on.

Bladder symptoms start as urgency and frequency, often nocturnal, and progress to urge incontinence with awareness; incontinence without awareness of the urge is not characteristic of iNPH and should point elsewhere ^[4].

Across series, gait disturbance occurs in 94% to 100% of patients, cognitive impairment in 78% to 98% and urinary dysfunction in 60% to 92%; 90.9% report urine dribbling and 74.5% incontinence ^[1]. Among 26 patients assessed neurologically, a snout reflex was found in 84%, an eyebrow reflex in 77%, paratonia in 65% and a palmar (palmomental) reflex in 61%; bradykinesia in 55%. Apathy was reported in 70.3% of patients, anxiety in 25% and wandering behaviour in about 10%. Detrusor overactivity is found in about 70% of those tested and bladder volumes are small at around 200 mL ^[1].

The general neurological examination is otherwise normal. Findings that are atypical and should redirect you toward other disease are lateralising signs, hemiparesis or paraparesis, spasticity, hyperreflexia and upper motor neuron signs; incontinence without awareness of urinary urge is also atypical ^[3]. Symptoms are symmetric: asymmetry implies another or additional diagnosis.

Severity is usually graded with the iNPH grading scale (0 to 4 for gait, cognition and incontinence, total up to 12) and with standard gait tests such as the timed up-and-go test, a short-distance straight walk and the Tinetti or Boon scales ^[1]. Screening cognition with the Mini-Mental State Examination or the Montreal Cognitive Assessment is advised; detailed neuropsychometry focuses on psychomotor speed, attention, working memory and executive function ^[4].

iNPH is rarely "pure". Most patients have comorbidities contributing to their symptoms, and other disorders can mimic the whole picture, so the first step is to exclude or treat what is treatable before testing for iNPH ^[4]. Important mimics include Alzheimer disease and other dementias, Parkinson disease and other parkinsonian disorders (progressive supranuclear palsy, dementia with Lewy bodies), vascular or multi-infarct dementia, cervical spondylotic myelopathy, lumbar spinal stenosis, normal-pressure hydrocephalus from a known cause (secondary), brain tumour or aqueductal stenosis (obstructive hydrocephalus), depression, and metabolic causes such as vitamin B12 deficiency and hypothyroidism. Delirium or a systemic illness in hospital is a frequent source of false attribution ^[3][4].

Gait impairment with urinary symptoms but no cognitive impairment should raise spinal cord and urological causes before invoking iNPH, and a patient with only dementia or only incontinence should be investigated for other disorders first, because in iNPH gait is almost always present and usually the presenting or worst symptom ^[3][4].

Neuroimaging is required for diagnosis; MRI is preferred. Lateral and third ventricles are enlarged and no obstruction to cerebrospinal fluid flow should be present. The Evans ratio, the widest frontal horn span divided by the widest brain diameter on the same axial slice, screens for ventriculomegaly: a ratio above 0.3 indicates large ventricles and above 0.33 very large ventricles, but this is not specific to iNPH ^[3][4]. DESH describes ventriculomegaly with effacement of the sulci over the high convexity and midline, often with enlarged Sylvian fissures and basal cisterns; a callosal angle of 90 degrees or less on a coronal image supports hydrocephalus rather than atrophy ^[4]. A pulsation artefact in the aqueduct or phase-contrast measurements of CSF flow cannot on their own justify surgery but support the diagnosis and the need for further testing ^[3]. Radionuclide cisternography has a high false-positive rate and is recommended against by both the international and Japanese guidelines ^[4].

A comparison of international and Japanese imaging practice puts the Evans ratio threshold at 0.3 in both guidelines, with research in healthy older adults suggesting that 0.33 may be more appropriate ^[4]. The callosal angle distinction uses a cut-off of 90 degrees with a reported sensitivity of 97%, specificity of 88% and positive predictive value of 93% for separating iNPH from Alzheimer disease ^[1].

Lumbar puncture also rules out other causes and can assess pressure. Most series and guidelines accept a CSF opening pressure of 200 mmH2O or less, although some series use 180 mmH2O; even higher values do not completely exclude iNPH, and there is no clearly defined lower limit in older adults ^[1]. CSF content is normally unremarkable. Biomarkers (amyloid-beta 42, tau proteins, neurofilament light chain, leucine-rich alpha-2 glycoprotein) can help distinguish iNPH from Alzheimer disease but no biomarker yet reliably predicts shunt response on its own ^[1].

The tap test should remove 30 to 50 mL through an 18- or 20-gauge needle, with a baseline gait examination immediately beforehand and a formal re-examination 2 to 4 hours later (some centres test at 30 minutes and again up to 24 hours), and patients may be allowed to remain active afterwards ^[3][4]. The infusion test requires specialist equipment and experience and is commoner in Europe than in North America ^[3][4]. Cerebrospinal fluid removal at 30 to 50 mL removes about the volume a shunt would divert in a day experimentally, and explains why gait change within hours is a reasonable predictor: in practice, 48 hours after the tap test gait improvement was seen, then declined ^[1].

A named practical point: there is no difference in tap-test performance whether 30 or 50 mL is removed ^[1]. ^[3]

The international guidelines (Relkin and colleagues, 2005, built on in Williams and Relkin, 2013) set three levels of diagnostic confidence: probable, possible and unlikely. The Japanese guidelines use a similar structure plus the probable requirement for DESH (third edition, 2021). The AAN practice guideline (2015) is not a full diagnostic guideline but grades the evidence for shunting and for predictors of response ^[2][4]. The Japanese and international criteria were compared directly by Williams and Relkin in table form: both require a compatible clinical picture and imaging, both use an Evans ratio of 0.3, and both accept CSF drainage testing, although only the international guideline requires or encourages infusion testing ^[4]. ^[3]

A few practical consensus statements run through all three:

• Gait impairment is typically the first or worst symptom, and a patient with only cognitive or only bladder symptoms needs evaluation for other disorders first ^[3][4].
• Onset is insidious over at least 3 to 6 months ^[3][4].
• Coexisting, treatable disease should be treated before shunt testing, because a resolved comorbidity may make shunt surgery unnecessary ^[4].
• No medical treatment, including acetazolamide, is supported by evidence as of 2013 to 2016 ^[3][4].

Management begins with confirming that no other disorder fully explains the symptoms and that any coexisting disorders are treated. Delirium must be resolved before testing, because it generates both false-positive and false-negative tap tests ^[4].

There is no medical therapy that improves iNPH. Acetazolamide is sometimes prescribed on the basis of reduced cerebrospinal fluid production in theory, but no evidence supports its use, and none of the guidelines recommends it ^[3][4]. Comorbid conditions (vascular risk factors, depression, parkinsonism, urological dysfunction) should be treated in their own right.

Shunt surgery is indicated for patients who improve after CSF drainage, or who have hydrodynamic variables consistent with iNPH. The international, Japanese and AAN sources all support shunt surgery as the effective treatment for properly selected patients ^[2][3][4]. Three Class III studies underpin the AAN conclusion: a 96% chance of subjective improvement and an 83% chance of improvement on a timed walk test six months after shunting, against 19% and 24% in non-shunted comparators ^[2]. In a randomised comparison, 32 of 49 patients in an immediate lumboperitoneal shunt group but only 2 of 44 in a delayed group improved by at least one modified Rankin Scale point at three months, a difference of 61% (95% confidence interval 42% to 68%) ^[2].

Endoscopic third ventriculostomy has not proved effective in iNPH and carries a higher mortality (3.2%) and complication rate (17.9%) than shunting (0.5% and 11.8%) in comparative data, with only 50% versus 76.9% symptomatic improvement at twelve months ^[1][3].

Ventriculoperitoneal (VP), ventriculoperitoneal with fixed or programmable valves, ventriculoatrial and lumboperitoneal (LP) shunts are all effective for iNPH ^[1]. The choice is made by the neurosurgeon and the patient. In Japan, a 2012 survey found LP shunts in 55.1%, VP shunts in 43.2% and ventriculoatrial shunts in 1.7% of treated patients ^[1]. No evidence supports one commercial valve over another, although most expert centres use programmable valves ^[3].

Programmable valves are recommended by the Japanese guideline because the pressure can be set to the individual and adjusted afterwards. Comparisons suggest that they reduce revisions and subdural collections relative to fixed-pressure valves, even though symptom improvement rates are similar ^[1]. The antisiphon or gravity-controlled devices further reduce overdrainage, although systematic comparisons have not separated their effects cleanly ^[1]. Initial pressure settings can be chosen from tables based on height and weight, and the Japanese guideline concludes that starting high and stepping down minimises overdrainage ^[1].

A practical point for every clinician: programmable valves can be altered by strong magnets held within 1 to 2 mm of the valve, including MRI, so the setting must be checked after imaging and corrected if necessary ^[3].

Patients should be seen regularly after surgery. Brain imaging is recommended in the first 6 to 12 months to exclude overdrainage, thin subdural effusions of 2 mm to 5 mm in an asymptomatic person usually not requiring a setting change ^[3]. Malfunction occurs in 30% or more of patients, often presents as gradual return of gait or cognition rather than an emergency, and, when identified and treated, about 75% of patients improve again ^[3]. Raising and lowering the valve setting for symptomatic overdrainage or underdrainage is the main advantage of programmable systems.

Complications divide into those of the disease and those of its treatment.

Untreated iNPH drives the triad forward: progressive gait dependence, cognitive decline, falls and loss of independence. Untreated patients become nursing-home dependent.

Shunt-related problems are the main serious burden. A pooled review cited by the AAN reported a mean complication rate of 38% and a 6% rate of permanent neurological deficit or death; another publication reported mortality for shunting of 5% to 15% ^[2]. The AAN guideline quotes an 11% serious adverse event risk in the pooled trial data and 22% in the SINPHONI trial ^[2]. Japanese survey data found overall surgical complication rates of 10% for VP and 14% for LP shunts, with no statistically significant difference; obstruction requiring revision was higher with LP (6.8% versus 1% with VP). Ventriculoatrial shunts gave more subdural haematomas than VP (12.7% versus 5.5%) though fewer occlusions (6.7% versus 15.3%) ^[1].

Specific complications to anticipate ^[1][2][3]:

• overdrainage with low-pressure headache worse on standing, nausea and, when severe, subdural effusion or haematoma
• shunt obstruction or disconnection
• shunt infection, ventriculitis or peritonitis, and, with lumbar catheter techniques, meningitis in 2% to 3% after external lumbar drainage
• cardiopulmonary complications, nephritis and renal failure after ventriculoatrial shunts, in less than 1% of cases ^[1]
• postoperative abdominal or shunt-related pain and, rarely, poor wound healing

Overdrainage is prevented and treated with programmable valves, antisiphon devices and serial setting changes rather than reoperation.

Properly selected patients have an 80% to 90% chance of responding to shunt surgery, and all three symptoms, including dementia, can improve ^[3]. The AAN summary is a 96% chance of subjective improvement and an 83% chance of improvement on a timed walk test at six months ^[2]. Japanese data give a more nuanced picture: improvement in 39% to 81% at three to six months, 63% to 84% at one year, 69% at two years and about 60% to 74% at three to six years, with decline in response over the very long term ^[1][2]. Gait usually improves most (60% to 77%), followed by cognition (61% to 69%) and urinary function (52%) ^[1]. Response measured on gait and timed walk tests is more robust than response measured on memory or reaction time, which deteriorated in one series from 83% at six months to 40% by five years ^[2].

Predictors of good response include a high outflow resistance on infusion testing, impaired cerebral blood flow reactivity to acetazolamide, a positive response to a tap test or external lumbar drainage, and a low comorbidity burden. Age by itself is not an independent predictor of poor response ^[1][2]. With a Kiefer comorbidity index of 3 or less, improvement was 93% at two years; above 4 points it was low, and at 6 points no improvement was expected ^[1]. Moderate to severe Alzheimer-type pathology is associated with a poor response. Short-term outcomes are influenced by complications, disease duration and severity, tap-test response and DESH; long-term outcomes are dominated by comorbidity ^[1].

When symptoms return after an initial response, about three-quarters of patients improve again if a shunt obstruction is found and corrected ^[3]. If comorbidity rather than obstruction is responsible, the situation needs to be recognised and managed on its own terms.

There is no primary prevention. The public-health challenge is finding the affected older adult: many with the condition never reach a hydrocephalus unit. Reasonable screening and early-action steps ^[3][4]:

• consider iNPH in dementia clinics, fall clinics and urology referrals for older adults with gait change
• investigate incidental ventriculomegaly on imaging, with re-evaluation every 1 to 2 years if the person is asymptomatic ^[3]
• treat other, more common causes without abandoning the iNPH work-up if symptoms do not resolve
• remember AVIM: in the Japanese series, 4 of 8 community-detected and 27 of 52 hospital-detected AVIM cases progressed (about 6 to 12% per year and about 17% per year respectively); these are rates from two separate cohorts, not one graded spectrum ^{[1] [2]}

Cost-effectiveness also argues for systematic case-finding, with shunting for iNPH described as medically and economically effective in the Japanese guideline, which estimated incremental cost-effectiveness ratios of 29,934 to 40,742 US dollars per quality-adjusted life year for a VP shunt and 58,346 to 80,392 for an LP shunt ^[1]. These figures are setting-specific.

Older adults. This is the core population. Frailty, vascular comorbidity and other degenerative disease confound both diagnosis and outcome, which is why the comorbidity indices were developed and why long-term results track comorbidity more than hydrodynamics ^[1][2]. A high comorbidity score should not automatically exclude shunting, because selected patients still benefit, but it must be discussed honestly with the patient and family.

Children. iNPH as defined here is a disorder of older adults. Ventriculomegaly in a child or young adult should be evaluated for obstructive or congenital hydrocephalus, and any communicating form in that age group usually follows an early insult and is managed as secondary hydrocephalus.

Pregnancy. iNPH is not a disease of pregnancy, and no specific guidance exists in the guidelines reviewed here; any hydrocephalus in pregnancy is managed by the neurosurgical and obstetric teams according to its cause.

A head CT establishing ventriculomegaly and a compatible clinical picture are enough to raise the diagnosis where MRI is unavailable; the Evans ratio and awareness of DESH are usable on CT. The tap test requires only a lumbar puncture and a careful gait assessment before and after, which is feasible in most settings. Infusion testing and external lumbar drainage are largely unavailable outside specialist centres; in their absence, a clearly positive tap test plus compatible imaging justifies referral for shunt surgery. Treatment requires neurosurgical expertise and a programmable valve is preferred but a fixed pressure valve is acceptable where that is all that is stocked. Refer patients early, because late presentation with established dementia responds less well.

• Normal pressure hydrocephalus is communicating hydrocephalus with large ventricles and normal or near-normal CSF pressure, usually in older adults.
• The clinical triad is gait and balance disturbance, cognitive decline with a frontal-subcortical pattern, and urinary urgency or incontinence; gait is usually first or worst and the full triad is present in only about 60%.
• Brain imaging shows ventriculomegaly; DESH, an Evans ratio of at least 0.3 and a callosal angle of 90 degrees or less support the diagnosis.
• Radionuclide cisternography is not recommended; tap testing with 30 to 50 mL and re-examination 2 to 4 hours later is the practical first diagnostic test.
• External lumbar drainage at about 10 mL per hour for about 72 hours and CSF infusion testing are the tests used in specialist centres.
• CSF pressure up to 200 mmH2O with normal content is the usual acceptable upper limit, but pressure alone never confirms or excludes the condition.
• Shunt surgery is the only effective treatment; medicines such as acetazolamide are not supported by evidence.
• Properly selected patients have an 80% to 90% chance of responding, with gait improving most and cognitive improvement lasting less well over years.
• Predictors of response include a high outflow resistance, impaired acetazolamide reactivity, positive drainage tests and a low comorbidity score; age alone is not a predictor.
• Shunt complications are common enough (roughly 10% to 38% depending on definition) that careful counselling, programmable valves and long-term review are essential.
• Malfunction occurs in 30% or more of patients but, when repaired, about 75% improve again.

• Cognitive impairment and dementia (frontal-subcortical syndromes)
• Parkinson disease and parkinsonian disorders
• Cerebrospinal fluid physiology and hydrocephalus
• Cervical spondylotic myelopathy
• Falls and gait disorders in older adults
• Urinary incontinence in older adults
• Vascular dementia
• Ventriculoperitoneal shunts