PRx · DERIVED

Pressure reactivity index (PRx)

A moving correlation between ICP and MAP at slow-wave frequencies, used to grade autoregulation continuously at the bedside and to derive an individualised CPP target.

ReactivityBedside + researchPeds + adultInvasiveEmerging

BLast reviewed 2026-05-1714-min read

5-minute summary

PRx is a continuous, moving Pearson correlation between mean ICP and mean MAP computed at slow-wave frequencies (~0.005–0.05 Hz, the Mayer-wave band).

Reactive vessels: When the cerebrovasculature is reactive (autoregulation intact), a small rise in MAP triggers vasoconstriction and ICP falls or stays the same: the correlation is negative or near zero (PRx ≈ −0.3 to 0).
Impaired autoregulation: When autoregulation is impaired, the vessels are passive pipes and ICP follows MAP: the correlation goes positive (PRx > +0.25).
CPPopt: Plotting PRx against CPP over the past 4 hours typically traces a U-curve; the vertex is CPPopt, the CPP at which the patient autoregulates best.
Evidence: PRx-guided CPP targeting (COGiTATE phase II) is the most evidence-supported individualised CPP strategy in severe TBI, with growing pediatric data (Tas 2022, 2024).
Requirements and limits: PRx requires high-quality continuous ICP and MAP signals and does not work in pulseless flow (ECMO, severe LV failure with non-pulsatile cardiac output) or when ICP signal quality is poor.

1. Bedside vignettes: why this matters in the PICU

Vignette A. Severe TBI day 2 in a 10-year-old, PRx says CPP is too low

A 10-year-old severe TBI day 2, post bifrontal decompressive craniectomy. ICP via parenchymal probe contralateral to the craniectomy reads 12 mmHg, CPP 60. Standard age-based protocol says CPP 50–60 is the floor and 60 is fine. The PRx trend over the past 4 hours has been +0.45 sustained. The CPPopt curve, fit to (CPP, PRx) data, gives a vertex at CPP 75 mmHg. The team lifts MAP with noradrenaline, CPP rises to 73, PRx falls to +0.05 within the hour. The default age-based target was inadequate for this child. Over the next 48 hours the team uses PRx-CPPopt as the operational CPP target and the patient's ICP-dose accumulates less than the matched cohort.

Vignette B. SAH day 5, PRx oscillating, real signal or artefact?

A 14-year-old with aneurysmal SAH post-coiling, day 5. PRx has been oscillating between 0.05 and +0.3 every 20 minutes for the past 4 hours. Is this fluctuating autoregulation (truly threshold-borderline) or is it artefact (motion, suctioning, sedation lightening)? The bedside team rejects 10-minute windows containing nursing care, looks at the clean overnight stretch when sedation was steady, and finds PRx +0.20 sustained. The CPPopt curve fits cleanly with a vertex at CPP 70. Lesson: PRx interpretation is window choice plus artefact rejection; you cannot trust a number computed across a suctioning episode.

Vignette C. Post-cardiac-arrest day 1, PRx not interpretable

A 7-year-old post-cardiac arrest day 1, targeted temperature 33°C. ICP-MAP correlation is near-zero with very low slow-wave power. The PRx number reads 0.05 but the underlying signal-to-noise is poor. The team holds off on PRx-guided targeting until day 3 when slow-wave power recovers, and uses COx (NIRS-MAP correlation) as the autoregulation surrogate in the interim. Lesson: PRx requires slow-wave content that is often absent in deep sedation, hypothermia, or non-pulsatile cardiac states. Have a fallback (Mx, COx, COx) for these scenarios.

2. What PRx is, and what it is not

PRx is a moving Pearson correlation coefficient between mean ICP and mean MAP, computed in successive overlapping windows.

\mathrm{PRx} = \mathrm{corr}(\mathrm{ICP}_{10\text{-s avg}}, \mathrm{MAP}_{10\text{-s avg}})

The standard implementation (Czosnyka 1997, refined by Smielewski's ICM+):

Sample ICP and MAP at 100–200 Hz.
Compute 10-second averages.
Take 30 consecutive 10-second averages (5 minutes of data).
Compute the Pearson correlation of the 30 paired (ICP, MAP) values.
Update every minute by sliding the window forward.

The logic. When autoregulation is intact, a rise in MAP triggers cerebrovascular vasoconstriction (smaller vessel volume, lower cerebral blood volume, lower ICP) within seconds. So MAP and ICP move in opposite or unrelated directions at slow-wave frequencies and the correlation is negative or near zero. When autoregulation is impaired, the vasculature is a passive pipe; MAP and ICP rise and fall together at slow frequencies and the correlation is positive.

Why slow waves. The autoregulatory response operates over seconds-to-minutes, in the same frequency band as Mayer waves (~0.1 Hz) and B waves (~0.02 Hz). Heart-rate frequency (~1 Hz) and respiration (~0.2 Hz) are filtered out by the 10-second averaging.

What PRx is not. It is not a direct measurement of autoregulation; it is a continuous index correlated with the cerebrovascular reactivity that defines autoregulation. It is not robust to artefact: motion, suctioning, transducer drift, and rapid sedation changes pollute the correlation. It is not interpretable in pulseless flow states (VA-ECMO with non-pulsatile output, severe LV failure) because the slow-wave drive disappears.

Clinical pearl

PRx is a signal, not a measurement. Trust trends over hours, not single 5-minute values. The CPPopt fit needs ~4 hours of data to be useful; isolated PRx values are noise.

In children

Pediatric slow-wave content differs. Younger children have higher heart rates and respiratory rates, shifting power in the slow-wave band. The standard adult window settings still work but signal-to-noise is sometimes lower. Pediatric validation studies (Tas 2022, 2024) used adult-style windowing and demonstrated the same CPPopt U-curve relationship with outcome. PRx > +0.25 is the same threshold used in pediatric work, but the CPPopt value will be lower (operational CPP floor 50 in a toddler vs 60 in an adolescent).

3. The math, in a deep-dive

3.1 Window choice and signal preparation

Sampling: ICP and MAP sampled at 100–200 Hz from the patient monitor.
Averaging: 10-second non-overlapping averages reduce high-frequency content (HR, respiration).
Window: 30 paired 10-second averages = 5 minutes.
Update rate: slide the window by 60 seconds, recompute Pearson.

3.2 Pearson correlation

For paired vectors $\vec{x}, \vec{y}$ of length $n$ :

r = \frac{\sum_{i=1}^{n}(x_i - \bar{x})(y_i - \bar{y})}{\sqrt{\sum_{i=1}^{n}(x_i - \bar{x})^2 \cdot \sum_{i=1}^{n}(y_i - \bar{y})^2}}

PRx is bounded $[-1, +1]$ .

3.3 Artefact rejection

ICM+ and other validated implementations reject windows when:

ICP variance is too low (flat trace, blocked catheter).
ICP variance is too high (catheter flush, suction artefact).
MAP variance is too low (arterial line damping).
The window contains nursing or procedural artefact flagged by clinician.

3.4 Why a 5-minute window

Short enough to be responsive to changes (suctioning, position) but long enough to capture slow-wave content. Shorter windows (1–2 minutes) are too noisy; longer (15+ minutes) lose temporal resolution.

3.5 The CPPopt fit

For each 5-minute window, plot the point (CPP, PRx). Accumulate ~4 hours of points. Bin CPP into 5-mmHg buckets, compute mean PRx in each bin. Fit a parabola; the vertex is CPPopt. The shaded "target band" is ±5 mmHg of CPPopt.

4. Thresholds: what counts as intact, impaired, and ambiguous

PRx range	Interpretation	Action
−0.3 to 0.0	Intact autoregulation	Continue current CPP target
0.0 to +0.25	Ambiguous / borderline	Lengthen window; check for artefact; re-evaluate in 30 min
> +0.25 sustained	Impaired autoregulation	Re-target CPP; check CPPopt curve
Falling over hours	Improving autoregulation	Consider weaning vasopressor support
Rising over hours	Worsening autoregulation	Investigate cause (ischaemia, oedema progression, seizures, sedation withdrawal)

The +0.25 threshold comes from a large body of outcome-association work, originally Steiner 2002 (mortality) and confirmed in many subsequent cohorts. Time spent with PRx > +0.25 correlates with 6-month outcome in adult severe TBI.

Caveat

A single PRx value is rarely actionable. The bedside team should look at a trend over the past 1–2 hours and ask "what changed?". A PRx jump from −0.1 to +0.3 over 30 minutes is more useful than a single +0.4 value.

5. The PRx trace and pattern library

Fig. 1

Four canonical PRx patterns over a 4-hour window. (a) Intact: PRx oscillating near zero or slightly negative, signal quality good. (b) Impaired: PRx sustained above 0.25 for hours, often paired with a flat or shallow CPPopt curve. (c) Oscillating: PRx swinging between 0 and 0.4 every 20-40 minutes, often reflecting fluctuating autoregulation threshold or transient sedation changes. (d) Artefactual: PRx jumping to extreme values then back, classic for movement, suctioning, transducer flush; should be rejected from the CPPopt fit.

MNM-Edu, original schematic.

Pattern	Bedside meaning	What to do
Flat near zero	Intact autoregulation	Continue current CPP
Sustained > +0.25	Impaired; needs CPPopt re-evaluation	Re-target CPP per CPPopt fit
Oscillating ±0.3	Threshold-borderline or sedation-fluctuant	Look for triggers; lengthen analysis window
Spikes to extremes	Artefact (motion, flush, suction)	Reject from CPPopt fit; verify with raw trace
Drifting positive over hours	Worsening autoregulation; new pathology	Investigate (seizures, oedema, ischaemia, sedation withdrawal)
Drifting negative over hours	Recovering autoregulation	Consider weaning interventions
Unstable / no signal	Low slow-wave power, pulseless flow, sedation depth	Switch to COx or Mx; check signal quality