PRx versus COx discordance

1. Three patient vignettes#

Vignette A. The septic TBI in a 10-year-old#

Lila, 10 years, 32 kg, severe TBI day 3 with ICP 14, MAP 75, CPP 61. PRx 4-hour average has been around 0.0 to 0.1 (intact) since admission. Day 3 she develops fever 39.2 C, lactate 3.2, blood culture positive for Gram-negative bacilli. Sepsis treatment initiated. Over the next 12 hours PRx remains 0.0 to 0.1 (intact); COx (from NIRS frontal probes) drifts from -0.1 to +0.4 (impaired). PbtO2 is 22 (acceptable). TCD MFV is unchanged. Clinical exam: unchanged sedation-restricted exam, NPi 4. Question: PRx and COx disagree; what is happening, and which to trust for CPP targeting? 1 2 3

Vignette B. The infant on ECMO#

Asher, 6 months, 7.5 kg, on VA-ECMO post-cardiac arrest, day 2. No ICP monitor (non-pulsatile circulation makes invasive ICP signal interpretation difficult). NIRS rSO2 is bilateral, 70% on the right and 55% on the left. COx-equivalent (NIRS-MAP correlation) over a 4-hour window is +0.5 on the left (impaired) and 0.0 on the right (intact). TCD shows non-pulsatile flow with HITS (high-intensity transient signals) cluster of 4 in the past hour on the right side. Question: what does the asymmetric NIRS plus the TCD HITS plus the COx asymmetry tell us, and what does it change? 4 5

Vignette C. The post-craniectomy patient with regional discordance#

Marisol, 14 years, 50 kg, severe TBI with right decompressive craniectomy day 4. The bone flap is off; the brain is exposed to atmospheric pressure on the right. ICP probe is in the left frontal parenchyma. PRx over the past 4 hours is +0.3 (impaired). Bilateral NIRS rSO2: right (under the craniectomy) 68%; left (intact bone) 64%. COx-equivalent on the right is -0.1 (intact); on the left is +0.4 (impaired). Question: the regional discordance reflects the regional anatomy; how do we use this information for CPP targeting, and what is the role of bilateral NIRS in the craniectomised patient? 6 7

2. The clinical question#

For each of these children: why do PRx and COx disagree, what does the discordance teach about the underlying physiology, and which index should drive CPP or MAP targeting decisions in this patient?

3. Pathophysiology refresher#

PRx and COx measure cerebral autoregulation through different physiological windows.

PRx (pressure reactivity index) is the moving Pearson correlation between ICP and MAP, computed at slow-wave frequencies (0.05 to 0.005 Hz; 5 to 10 second averages over a 5 to 30 minute window). The mechanism: in an autoregulated brain, an increase in MAP triggers cerebrovascular constriction, which reduces cerebral blood volume, which reduces ICP. Hence MAP and ICP are negatively correlated (PRx < 0). In a disautoregulated brain, MAP increases produce CBV increases (passive flow), and ICP rises with MAP (PRx > 0). PRx is the most validated autoregulation index in adult TBI. 3

COx (cerebral oximetry index) is the moving correlation between NIRS rSO2 and MAP. The mechanism: in an autoregulated brain, MAP increases drive cerebrovascular constriction in the same way PRx assumes, which keeps tissue oxygen delivery roughly constant; rSO2 stays flat. In a disautoregulated brain, MAP increases drive passive flow rises and tissue oxygenation increases; rSO2 tracks MAP. So COx > 0 (positive correlation) signals impaired autoregulation. COx is the non-invasive analogue of PRx, with the advantage of being available at the tissue level via NIRS. 1 2

Why can they disagree?

• PRx is global (one ICP signal averaged over the intracranial space); COx is regional (one or two NIRS probes over specific cortex). Regional autoregulation can differ from global autoregulation in focal injury, after craniectomy, or with regional sepsis-driven shunting.
• Sepsis-driven microvascular shunting. In severe sepsis, the microvascular bed develops shunt physiology: arterioles dilate independently of metabolic demand, and oxygen is delivered to tissue regions that do not use it while other regions are under-supplied. The macrovascular response (PRx) may remain intact because the proximal vessels still respond to MAP; the microvascular response (COx) becomes impaired because the autoregulatory linkage between metabolic demand and tissue oxygenation is broken. PRx intact, COx impaired is the classical sepsis signature.
• NIRS scalp and extracranial contamination. NIRS samples a path that includes scalp, skull, and superficial cortex. Up to 30% of the signal in some configurations comes from extracranial tissue. Scalp perfusion is autoregulated differently from intracranial perfusion, and contamination distorts the COx signal. PRx is unaffected. 8
• Probe placement. PbtO2-based autoregulation indices (Px) depend on which tissue is being sampled (peri-contusional, distal-from-injury, healthy contralateral). Px in injured tissue may be impaired while a global index like PRx is intact. NIRS placement matters similarly: an NIRS probe over an unaffected region may show intact COx while the rest of the brain is impaired.
• Time scales. PRx and COx integrate over different windows; transient changes in MAP may affect one before the other.
• Calibration and signal quality. PRx requires a stable ICP signal; COx requires a clean NIRS signal. Both degrade with motion artefact, electrode drift, or sensor displacement.

Why does it matter clinically? When PRx and COx disagree, the underlying physiology is informative. Sepsis-driven shunting is a real phenomenon, not an artefact, and it changes management (the macrovascular CPP target may be met while the tissue is still oxygen-distressed). Probe placement matters; regional discordance after craniectomy reflects regional anatomy. NIRS contamination is technical and can be mitigated. The clinical task is to interpret the disagreement, not dismiss it.

4. The multimodal picture#

5. Decision tree#

flowchart TD
  PRxCOxDisc[PRx and COx disagree] --> Sepsis{Sepsis present?}
  Sepsis -->|Yes| Shunt[Likely microvascular shunting; trust PRx for CPP; address sepsis]
  Sepsis -->|No| Scalp{NIRS probe quality good?}
  Scalp -->|No| Recheck[Recheck probe contact, reapply]
  Scalp -->|Yes| Regional{Focal injury or craniectomy?}
  Regional -->|Yes| Regdiff[Regional autoregulation differs; interpret per probe location]
  Regional -->|No| Mx{TCD Mx available?}
  Mx -->|Yes| Tiebreak[Use Mx as tiebreaker]
  Mx -->|No| Default[Fall back to default CPP thresholds plus clinical exam]
  Shunt --> Action[Treat sepsis; reassess in 6 to 12 h]
  Regdiff --> Action
  Tiebreak --> Action
  Default --> Action
  Action --> Followup[Recheck both indices q 4 h]

6. Step-by-step bedside actions#

For Lila (10 y, 32 kg, septic TBI with PRx intact, COx impaired). Times are from sepsis onset.

1. 0 to 30 min: confirm the discordance. Re-check NIRS probe contact, reapply if any loose pad. Verify ICP transducer zero. Recompute PRx and COx in the next 5-minute window with fresh signal. If discordance persists, it is real.
2. 30 to 60 min: address the sepsis. Source control (cultures, empirical antibiotics already started, lactate trend, fluid responsiveness assessment, vasopressor titration to MAP target). The most likely mechanism is sepsis-driven microvascular shunting; treating the sepsis is the primary intervention.
3. 60 to 90 min: which index drives CPP for the next 6 to 12 hours? Default to PRx (more validated, less contaminated by sepsis-related microvascular changes) for the CPP / MAPopt target. Monitor COx as a secondary signal.
4. 60 to 90 min: cross-check with TCD Mx. If available, TCD-based Mx provides a non-invasive macrovascular index that should agree with PRx in pure macrovascular autoregulation. Mx agreement with PRx supports the macrovascular-intact interpretation.
5. 60 to 120 min: PbtO2 if available. PbtO2 less than 20 mmHg in this physiology adds tissue-level evidence; PbtO2 acceptable with PRx intact and COx impaired supports the microvascular-shunting interpretation rather than true autoregulation failure.
6. 6 to 12 h: reassess. With sepsis treatment, COx should recover if microvascular shunting is the mechanism. Persistent COx impairment after 12 h of effective sepsis treatment suggests an alternative driver (probe contamination, regional injury, true autoregulation failure progressing).
7. NIRS technical review. If discordance persists, check for probe drift, oedema under the probe, hair interference, lighting interference. Re-position probes if necessary.
8. Document and hand over. Brief the next shift on the discordance interpretation and the chosen CPP target.
9. If CPP target shifts: raise CPP by 5 mmHg if COx and clinical exam jointly suggest under-perfusion despite PRx-intact macrovascular response. Avoid large MAP excursions; small stepwise changes.
10. Long-term: the PRx-COx discordance during sepsis is a known phenomenon; document its occurrence and resolution in the patient summary.

7. Management endpoints#

Success looks like: discordance resolves with sepsis treatment; COx returns toward 0.0 to -0.1; PRx remains stable; CPP is in the targeted range; PbtO2 acceptable; clinical exam stable.

Failure looks like: persistent COx impairment despite sepsis treatment; PRx degrades (autoregulation now globally impaired); clinical deterioration; rising ICP; falling PbtO2.

When to escalate:

• New global autoregulation failure (PRx degrades), increase ICP-directed therapy; consider osmotherapy, sedation deepening, hyperventilation cautiously.
• Persistent tissue-level oxygen distress despite acceptable CPP and PRx, transfusion if anaemic, raise FiO2 if responsive, consider PbtO2-directed CPP raise.
• Refractory shock that prevents CPP targeting, address haemodynamics primarily.

When to de-escalate:

• Both indices recover toward intact range.
• Sepsis under control with declining lactate and afebrile.
• Stable haemodynamics with weaning vasopressor.
• Clinical exam improving.

8. Variant subsections#

8.1 PRx-Mx discordance (TCD versus invasive)#

PRx is global ICP-MAP; Mx is the TCD-MFV-MAP correlation. In intact macrovascular autoregulation, the two should largely agree. Discordance arises when (a) TCD signal quality is poor; (b) the TCD probe is sampling a different region than the ICP transducer represents; (c) the patient has proximal vessel stenosis or vasospasm that affects MFV without affecting ICP. The interpretation: PRx is the gold-standard macrovascular index when available; Mx is the non-invasive fallback. 3

8.2 COx-Mx discordance (NIRS versus TCD, both non-invasive)#

In the absence of invasive ICP, COx and Mx are the two non-invasive autoregulation indices. They sample different windows (tissue versus macrovascular flow) and can disagree. The interpretation: COx is more sensitive to microvascular changes; Mx is more sensitive to macrovascular changes. Both impaired = global autoregulation failure; COx impaired only = microvascular failure with macrovascular preservation (sepsis pattern); Mx impaired only = macrovascular failure (rare in pure form; usually accompanies global failure).

8.3 PRx-Px discordance (regional PbtO2)#

Px uses PbtO2-CPP correlation; PbtO2 is highly local (within a few millimetres of the probe tip). Px in peri-contusional tissue may be impaired while global PRx is intact. This is a known phenomenon in adult TBI literature. The interpretation: regional Px informs regional CPP targeting (e.g., raise CPP further to support the peri-contusional region); global PRx informs the overall MAPopt target. 9 10

8.4 PRx-COx discordance in HIE post-arrest#

In post-cardiac-arrest physiology, PRx and COx may both be impaired during the early hyperaemic phase; both may recover or both may further degrade depending on injury severity. Discordance in this setting may reflect probe placement asymmetry (one NIRS over more-injured tissue) or post-arrest microvascular dysfunction. 11 12

8.5 Post-craniectomy regional discordance#

Decompressive craniectomy removes the bone over the affected region. The brain on the craniectomy side is exposed to atmospheric pressure (no skull-vault pressure containment); ICP measured by a left-sided probe represents the contralateral hemisphere. NIRS over the craniectomy side and the intact side will often differ substantially in COx because the underlying physiology is different. Regional interpretation is essential. 7

8.6 COx in resource-limited settings#

In centres without invasive ICP, COx is one of the non-invasive autoregulation indices that can drive MAP targeting. The validation base in paediatric populations is smaller than in adult populations; provisional use with conservative thresholds is the current state. Combination with clinical exam and Mx (where TCD available) increases confidence. 13 14

9. Multimodal integration matrix#

10. Worked alternative scenarios#

10.1 What if PRx and COx both move together?#

Concordant indices are the simpler case. Both intact = autoregulation is intact, CPP at the current level is acceptable. Both impaired = autoregulation is broken, the CPP threshold is more critical (defer to default thresholds; investigate cause; aim for narrower MAP control). Concordance does not mean perfect; it means the macrovascular and microvascular responses are matched.

10.2 What if PRx is intact but COx asymmetric?#

In a patient with focal injury (contusion, stroke), bilateral NIRS may show one hemisphere with impaired COx and the other intact. Global PRx may be intact because the unaffected hemisphere dominates the global signal. The interpretation: regional injury produces regional autoregulation failure; consider regional therapeutic targeting (raise CPP modestly to support the affected region) and watch for clinical decline.

10.3 What if the NIRS probe is over scalp haematoma?#

Scalp haematoma (post-trauma) under a NIRS probe produces aberrant rSO2 readings and unreliable COx. The clinical clue: localised scalp swelling visible on exam; rSO2 with patterns unrelated to systemic physiology; COx erratic. Solution: move the NIRS probe to an unaffected scalp location or to the contralateral side.

11. Outcome data#

• Brady 2010 COx (paediatric ECMO): original paediatric COx validation showed that NIRS-MAP correlation tracks autoregulation status on ECMO; COx greater than 0.3 identified periods of impaired autoregulation. 1
• Lee 2009 non-invasive autoregulation (paediatric): demonstrates that NIRS-derived indices can identify autoregulation status non-invasively in children. 2
• Rivera-Lara 2017 autoregulation review: comprehensive review of indices, including PRx, COx, Mx, Px, and their respective strengths and weaknesses; emphasises that no single index is gold standard in all populations. 3
• Oddo 2017 microdialysis and autoregulation: integration of tissue-level monitoring with autoregulation indices; emphasises the value of multimodal interpretation. 6
• Andresen 2014 NIRS in TBI: NIRS limitations and the scalp-contamination problem; addresses the technical sources of COx-PRx discordance. 7
• Davies 2017 NIRS in acute injury: comprehensive review of NIRS interpretation including the signal-contamination issue. 8
• BOOST-II (Okonkwo 2017): PbtO2-directed CPP management trial in adult TBI; demonstrates that regional tissue oxygen targeting can improve outcomes; relevant to Px-PRx discordance. 9
• BOOST-3 (Bernard 2025): larger trial of PbtO2-directed care building on BOOST-II. 15

12. Pitfalls#

• Treating discordance as error. PRx-COx discordance is information about physiology, not noise.
• Ignoring sepsis as a driver. Sepsis-driven microvascular shunting is a known cause of PRx-COx discordance; treat the sepsis.
• Believing COx is "just non-invasive PRx". They measure different physiology; they are complementary, not interchangeable.
• Forgetting scalp contamination of NIRS. Up to 30% of the NIRS signal can be extracranial; this affects COx without affecting PRx.
• Single-probe NIRS in focal injury. Bilateral NIRS is essential when injury is asymmetric; a single probe misses regional discordance.
• Using PbtO2-derived Px as a global index. Px is local; do not extrapolate to the whole brain.
• Aggressive CPP changes based on a single 5-minute window. Both PRx and COx require longer windows for reliable interpretation; small stepwise changes with re-assessment are safer than large excursions.
• Forgetting clinical exam. When indices disagree, clinical exam is the tiebreaker.

13. Pediatric considerations#

14. Combine with#

• NIRS modality page: rSO2 interpretation, scalp contamination, COx derivation.
• PRx modality page: macrovascular autoregulation, CPPopt.
• TCD / TCCD modality page: Mx, the TCD-based autoregulation index.
• PbtO2 modality page: tissue oxygen, BOOST-II and BOOST-III.
• Integration: CPPopt targeting: the U-curve approach to CPP titration.
• Integration: PbtO2-CPP titration: the BOOST-style management ladder.
• Integration: Discordance triage: how to triage when multimodal monitors disagree.
• Foundations: autoregulation: the Lassen curve, the limits, the indices.

15. Evidence summary#

16. Recent literature (2022 to 2025)#

• BOOST-3 (Bernard 2025) continues the BOOST-II line of evidence on PbtO2-directed CPP management; informs the integration of Px with PRx. 15
• Helbok 2024 paediatric MMM update addresses the role of multiple autoregulation indices in paediatric populations. 14
• Figaji 2025 paediatric MMM consensus formalises the use of autoregulation indices in resource-stratified bundles. 13
• Naim 2023 PCCM addresses the post-arrest physiology where PRx-COx discordance is informative. 12
• TCD-based Mx continues to be validated in paediatric populations; robotic TCD platforms have made continuous Mx more feasible.
• NIRS scalp contamination remains an active area of research; algorithms for extracranial subtraction are improving but not yet standard.

17. Self-check#