A sham-controlled trial finds brain shunts help walking, not memory

Idiopathic normal-pressure hydrocephalus is defined by a triad — impaired gait and balance, declining cognition, and loss of bladder control in older adults — and treated by implanting a shunt to drain cerebrospinal fluid. The treatment has been in use for decades. Whether it works has never been tested against a sham.

The PENS trial, published in the New England Journal of Medicine on 16 September, ran that test, and the answer separates the triad [s1]. Shunting improved walking. It did not improve memory or continence.

The design solves an old problem

Sham surgery trials are ethically fraught and logistically hard. PENS sidestepped part of the difficulty using a device feature: every participant received a real, noninvasively adjustable shunt, and the randomisation was to the valve setting rather than to surgery itself [s1].

Participants were assigned to an open-shunt valve setting with an opening pressure of 110 mm of water, or a placebo setting with an opening pressure above 400 mm of water — high enough that essentially no cerebrospinal fluid drains [s1]. The trial was double-blind [s1].

Everyone was operated on. Everyone had hardware. What differed was whether the hardware did anything. That removes the surgical experience, the anaesthesia and the recovery period from the comparison, leaving only the drainage.

Participants had been selected for surgery on the basis of gait-velocity improvement with temporary cerebrospinal fluid drainage [s1] — the standard clinical screening test. This means the trial enrolled the patients most likely to respond, and its results should be read as applying to that selected group rather than to everyone carrying the diagnosis.

The primary outcome was change in gait velocity three months after surgery [s1]. Secondary outcomes at three months were the Tinetti scale total score (range 0 to 28; lower scores indicate worse gait and balance), the Montreal Cognitive Assessment (range 0 to 30; lower is worse cognition), and the Overactive Bladder Questionnaire (range 0 to 100; higher is worse incontinence) [s1].

The result

Ninety-nine participants underwent randomisation and received the assigned intervention [s1].

At three months, gait velocity increased by a mean of 0.23 ± 0.23 m per second in the open-shunt group, assessed in 49 participants, and was unchanged in the placebo group at 0.03 ± 0.23 m per second, also assessed in 49 — a treatment difference of 0.21 m per second (95% CI, 0.12 to 0.31; P<0.001) [s1].

The Tinetti score improved by 2.9 points with an open shunt versus 0.5 with the placebo setting (P=0.003) [s1]. The other two secondary outcomes did not separate meaningfully: MoCA changed by 1.3 points versus 0.3, and the Overactive Bladder Questionnaire by −3.3 versus −1.5 [s1].

The harms are not trivial

The adverse event results were mixed, and they point in both directions [s1].

More participants in the placebo group reported falls, 46% versus 24% — consistent with the gait benefit being real and having a downstream consequence [s1]. Cerebral bleeding occurred in 2% of each group [s1].

But more participants in the open-shunt group had subdural bleeding, 12% versus 2%, and positional headaches, 59% versus 28% [s1]. Both are recognised consequences of over-drainage. A one-in-eight rate of subdural bleeding over three months is a substantial harm to set against a 0.21 m/s gait improvement, and the trial does not weigh them for the reader.

How this changes the prior evidence

The only previous randomised trial of any size compared timing rather than function. SINPHONI-2, an open-label trial at 20 Japanese centres, randomised 93 patients aged 60 to 85 with idiopathic normal-pressure hydrocephalus to lumboperitoneal shunt surgery within one month or surgery postponed for three months [s2]. Patients and assessors were not masked [s2].

At three months, 32 of 49 (65%) in the immediate group had improved by at least one point on the modified Rankin scale, versus 2 of 44 (5%) in the postponed group — a 61-point difference (95% CI, 42 to 68; p<0.0001) [s2]. But at 12 months after surgery, once both groups had been operated on, the proportions were similar: 30 of 45 (67%) versus 22 of 38 (58%), a difference of 9 points (95% CI, −14 to 31; p=0.496) [s2].

An unmasked trial comparing surgery now against surgery later cannot distinguish drainage from expectation. PENS can, and it confirms that a real effect exists — while narrowing what that effect consists of.

What is not established

The trial's primary endpoint was measured at three months [s1]. Idiopathic normal-pressure hydrocephalus is a chronic condition managed over years, and nothing here speaks to whether the gait benefit persists, or whether the subdural bleeding rate rises with longer drainage.

Ninety-nine participants is small, which is why several of the secondary comparisons are reported without confidence intervals or p-values in the abstract, and why the cognition and continence results are best read as "no benefit demonstrated" rather than "benefit excluded."

And the selection criterion matters. These were patients whose gait improved with temporary drainage [s1]. The trial does not tell clinicians what to do about patients who fail that screen, which is a substantial share of those who carry the diagnosis.

The trial was funded by the National Institute of Neurological Disorders and Stroke and the Trial Innovation Network, and registered as NCT05081128 [s1].

This article is informational and is not medical advice.

Sources

• [s1] "A Randomized Trial of Shunting for Idiopathic Normal-Pressure Hydrocephalus," New England Journal of Medicine, published online 16 September 2025. https://doi.org/10.1056/NEJMoa2503109
• [s2] "Lumboperitoneal shunt surgery for idiopathic normal pressure hydrocephalus (SINPHONI-2): an open-label randomised trial," The Lancet Neurology, 28 April 2015. https://doi.org/10.1016/S1474-4422(15)00046-0