COx, the NIRS-based autoregulation index

1. Bedside vignettes: why this matters in the PICU#

Vignette A. The preterm infant on SafeBoosC-style monitoring#

A 26-week preterm infant on day 2 of life has continuous bilateral NIRS pads on the frontal cortex (rSO2 target 55 to 85% per SafeBoosC protocol). MAP is 38 mmHg on a low-dose dopamine infusion for borderline hypotension. COx (rolling 5 min) is +0.4, sustained over the last 6 hours; rSO2 is at the lower end of target (58%) and trending down. Autoregulation is impaired and tissue oxygenation is borderline. Action: per SafeBoosC framework, increase MAP target gently with volume or vasopressor titration; reassess COx and rSO2 every 30 min; do not push too high (the upper bound matters in the preterm brain). 4 3 5

Vignette B. Post-cardiac-surgery toddler with discordant COx and clinical exam#

An 18-month-old on day 1 after a complex Norwood stage 1 repair, sedated, ventilated. NIRS rSO2 reads 70% bilateral, MAP 55 mmHg, COx = +0.5 sustained, but ICP is not monitored. The bedside platform plots COx-vs-MAP across the last 4 hours: U-curve vertex at MAP 60 (MAPopt 60). Action: gentle MAP escalation (volume or noradrenaline) toward MAP 58 to 63; reassess COx in 30 min. The COx framework here lets the team derive an individualised MAP target without an invasive ICP monitor or sustained TCD. 1 6

Vignette C. The septic teenager with discordant COx and PRx#

A 15-year-old with septic shock, intubated and sedated, MAP 72 on noradrenaline. ICP monitor in place (placed earlier for confused septic encephalopathy and headache). PRx (bedside ICM+) reads −0.1 (intact); COx reads +0.4 (impaired). The neuropals team discusses: PRx (macrovascular ICP-based) suggests intact autoregulation; COx (tissue-level NIRS) suggests impaired microvascular function consistent with septic microvascular shunting. The discordance is itself informative: it argues for sepsis-specific management (source control, optimisation of oxygen delivery and utilisation) rather than aggressive MAP escalation. 7 1

2. What COx is, and what it is not#

COx is a moving-window Pearson correlation between the NIRS-derived regional cerebral oxygen saturation (rSO2, %) and the mean arterial pressure (MAP, mmHg). Computed over the slow-wave band (typically 0.003 to 0.05 Hz) using 10-second averages over a 5-minute rolling window, and updated every 1 minute.

Three concepts to anchor.

COx is a tissue-level autoregulation index. Where PRx uses ICP (an integrated whole-brain pressure signal) and Mx uses TCD MFV (a large-vessel velocity signal), COx uses rSO2 (a tissue-compartment oxygen signal). The three indices share a common physiology (slow-wave correlation captures autoregulatory coupling) but they probe different compartments. 1 2

COx is the easiest autoregulation index in pediatrics. NIRS pads are non-invasive, cheap, and widely available in NICU and pediatric ICU. No bolt, no TCD probe, no headframe. The trade-off: rSO2 is a smaller signal than ICP or MFV, more easily corrupted by movement / sweat / hair / extracranial contamination, and reflects only the frontal cortex ~2 to 3 cm deep.

COx interpretation thresholds:

Like PRx and Mx, COx generates an COx-vs-MAP U-curve from which MAPopt can be derived. 8 9

3. The signal: COx vs PRx vs Mx#

Illustrative teaching values (MNM-Edu schematic), not a single cited cohort. Neonatal cerebral autoregulation is monitored with MAP-based NIRS indices (COx and the haemoglobin volume index, HVx), not ORx, which needs invasive CPP/PbtO2 rarely available in neonates. Optimal MAP clusters approximately 50 to 60 mmHg in term cooled HIE and tracks lower (roughly gestational-age-based) in preterms, where cohort data are sparse. COx, HVx and TOHRx are not numerically interchangeable and no standardised threshold exists; approximately 0.3 is the most common cutoff. Sparse

COx and ORx are related but distinct indices: COx correlates MAP with NIRS rSO2 and is fully non-invasive, the practical choice when no ICP is in place (as in most neonates), whereas ORx correlates CPP (which requires an ICP monitor) with brain-tissue oxygenation or rSO2 and is used in adults who already have ICP.

The three indices have complementary strengths:

The three can disagree. The most studied disagreement is PRx-COx discordance in sepsis: PRx (intact, macrovascular) while COx (impaired, microvascular) suggests sepsis-related microvascular shunting where the large vessels autoregulate but tissue oxygenation does not. This is informative, not error. 7

4. The signal: COx computation in detail#

The bedside platform requires:

1. Continuous NIRS rSO2 (typically 1 to 10 Hz sample rate, frontal cortex pad).
2. Continuous arterial line MAP (100 Hz minimum).
3. Time-synchronisation between NIRS device and arterial line.
4. Slow-wave extraction: band-pass filter (0.003 to 0.05 Hz, or 0.01 to 0.05 in neonates with very rapid respiration) on both signals.
5. Rolling Pearson correlation: 10 s averages over 5 min window, updated every 1 min.
6. COx-vs-MAP binning: 4 to 6 hours of data; bin MAP into 5 mmHg windows; mean COx per bin; parabolic fit; vertex = MAPopt.
7. Display: continuous COx trend, MAPopt U-curve, ±5 mmHg target band.

Variants in the NIRS autoregulation family:

• THx (tissue haemoglobin reactivity): uses NIRS-derived tissue haemoglobin index instead of rSO2.
• HVx (haemoglobin volume index): uses NIRS-derived oxyHb or total Hb against MAP.

These are conceptually related to COx; the bedside platform (ICM+, Sickbay) typically supports several variants. 1 2

5. The numbers: what to record at the bedside#

Display COx alongside rSO2 trend, MAP, and (where present) PRx, Mx. The multi-index multimodal view is the modern standard.

6. What is normal? COx interpretation reference#

Pediatric and neonatal data: SafeBoosC-related studies and Brady piglet work support these thresholds; the absolute values are similar to adult. 8 1 7

7. What is abnormal? Pattern library#

Decision tree: "what does COx tell me?"#

flowchart TD
  COx[COx reading] --> Q1{COx > +0.3 sustained?}
  Q1 -->|No| OK[Intact; continue MAP target]
  Q1 -->|Yes| Q2{NIRS signal quality good?}
  Q2 -->|No| Fix[Re-prep probe; pause interpretation]
  Q2 -->|Yes| Q3{PRx or Mx available?}
  Q3 -->|Yes, concordant| Conc[Concordant impaired; narrow MAP range]
  Q3 -->|Yes, discordant| Disc[Discordant; investigate microvascular]
  Q3 -->|No| Act[Treat as impaired; tighter MAP control]

8. Try it: interactive widgets#

9. COx-driven management decisions#

9.1 Preterm and SafeBoosC framework#

The flagship pediatric / neonatal autoregulation context. SafeBoosC-II showed feasibility of NIRS-guided rSO2 targeting in preterm infants; SafeBoosC-III (Plomgaard 2024, Hansen 2023) did not demonstrate mortality benefit at 2 years but the framework (rSO2 target band, COx-derived MAPopt support) remains established practice in many neonatal units. 4 5 3

9.2 HIE and post-cardiac arrest#

In neonatal HIE, COx (or related NIRS-derived autoregulation indices) provides bedside autoregulation status during the rewarming window and the days that follow. Impaired COx is associated with worse neurodevelopmental outcome in some cohorts; the data are evolving. 2 6 10

9.3 Post-cardiac-surgery (CHD)#

The post-Norwood / arterial switch / Glenn / Fontan patient is at high risk for cerebral injury. Bilateral NIRS pads are routine; COx adds the autoregulation read; MAP targeting can be individualised by COx-MAPopt in selected centres. 6

9.4 Pediatric ECMO#

COx is well-suited to ECMO because NIRS pads work on the non-pulsatile circulation (where TCD-derived Mx struggles). The ELSO neurological surveillance bundle includes bilateral NIRS; COx is an emerging adjunct. 11 12 13

9.5 Sepsis and microvascular shunting#

COx-PRx discordance (COx impaired, PRx intact) is a research-grade signal of microvascular shunting in sepsis. Management implications: avoid aggressive MAP escalation (which does not improve tissue oxygenation in this state); focus on source control, oxygen delivery (Hb, SaO2), and oxygen utilisation (mitochondrial support). 1 7

9.6 Pre- and post-monitor windows in TBI / SAH#

In a TBI patient awaiting a bolt or after bolt removal, COx provides continuity of autoregulation-guided care, similar to Mx but using NIRS instead of TCD. Particularly useful in centres without TCD capability. 14 15

10. Clinical contexts: COx across acute brain injuries#

10.1 Severe TBI (pediatric or adult)#

Less common indication than PRx or Mx but useful in selected scenarios (pre-bolt placement, post-bolt removal, centres without invasive monitoring). Adult COx data exist (Brady's group, Highland 2014); pediatric data sparse. 1 14

10.2 Aneurysmal SAH#

COx is less validated in SAH than PRx or Mx but is feasible. Particularly relevant post-monitor-removal or when bedside TCD is intermittent. 15 16

10.3 Pediatric AIS#

NIRS is part of the post-recanalisation pediatric AIS monitoring bundle; COx is an emerging research adjunct. 17 18

10.4 HIE and post-cardiac arrest#

COx (and related NIRS autoregulation indices) features in pediatric HIE cohort studies. Lee 2009 reported impaired NIRS-derived autoregulation correlating with worse neurodevelopmental outcome. 2 10 6 19

10.5 Pediatric ECMO#

A natural fit because NIRS pads work on non-pulsatile circulation. ELSO neurological guidelines incorporate bilateral NIRS; COx is the emerging autoregulation adjunct. 11 12

10.6 Meningitis and encephalitis#

Less established. NIRS adds tissue oxygenation surveillance; COx is an investigational add-on. 20 21

10.7 Brain-death determination#

Not a brain-death tool. As cerebral circulatory arrest evolves, rSO2 may fall and COx may degenerate; the formal diagnosis remains clinical + apnoea + ancillary per local protocol. 22

10.8 DKA cerebral oedema#

Limited published COx use. Investigational. 23 24

10.9 Sepsis and septic encephalopathy#

The most physiologically interesting COx context: microvascular shunting can decouple tissue rSO2 from systemic MAP, manifesting as impaired COx with intact PRx. Research interest in using this pattern as a sepsis-specific bedside marker. 1 7

10.10 Preterm neonates (SafeBoosC framework)#

The leading pediatric NIRS autoregulation context. SafeBoosC-II and -III trials, the field-defining studies. Framework: target rSO2 within a band (55 to 85%); use COx-derived MAPopt as a refinement when reliable. 4 5 3

11. Multimodal integration: COx in the MMM/MNM stack#

12. Setup and technique#

12.1 Equipment#

• NIRS device: INVOS, FORE-SIGHT, NIRO, or equivalent; bilateral frontal cortex pads.
• Synchronised arterial line MAP: 100 Hz minimum.
• Bedside platform: ICM+, Sickbay, custom Python pipeline, or vendor-integrated COx computation.
• Skin preparation: clean and dry forehead; remove hair if needed; some pads have integrated adhesive.
• Quiet environment: head movement, scalp pressure changes, and ambient bright light all corrupt NIRS.

12.2 The setup workflow#

1. Place bilateral NIRS pads on the frontal cortex (typically over Fp1 and Fp2 of the 10-20 system).
2. Verify good signal quality: rSO2 should read stably within 10 seconds; signal-strength indicator should be in the high range.
3. Confirm MAP recording: calibrated, square-wave test passed.
4. Confirm time-synchronisation.
5. Start the rolling COx computation (5 min window, 1 min update).
6. Display COx, rSO2 (both sides), MAP, and any paired indices (PRx, Mx).

12.3 The COx-MAPopt fit#

1. Collect ≥ 4 hours of data.
2. Bin MAP into 5 mmHg windows; compute mean COx per bin.
3. Fit a parabola.
4. Vertex = MAPopt.
5. Target MAP within ±5 mmHg of MAPopt.
6. Re-fit every 1 to 4 hours.

12.4 Quality control#

• Pad contact: re-secure or re-prep every 12 to 24 h; signal quality drifts.
• Scalp blood flow contamination: a common confounder; the NIRS algorithms attempt to subtract but are imperfect.
• Sweating and hair: cause signal degradation; clean and re-secure as needed.
• Movement artefact: in unsedated children especially; flag and exclude affected epochs.
• Document quality with every COx report.

12.5 Pediatric / neonatal specific tips#

• Small pads for neonates; the larger adult pads do not contact well on small foreheads.
• Fontanelle: open fontanelle alters depth of NIRS signal; calibrate within centre.
• Scalp blood flow is high in neonates; extracranial contamination is a recognised confounder.
• SafeBoosC training: NICUs running SafeBoosC-style protocols typically have formal NIRS training; COx is an emerging adjunct.

12.6 When COx is not the right tool#

• Persistent poor signal quality: hair, sweat, head trauma, scalp burn; pause interpretation.
• Inadequate MAP variation: no MAPopt fit possible.
• Heavy scalp contamination: rSO2 dominated by extracranial blood flow; cortical signal lost.
• Pre-arrest state: rSO2 falls precipitously; COx becomes uninterpretable.

13. Pitfalls#

• Extracranial contamination: scalp blood flow contributes to rSO2; COx may reflect peripheral haemodynamics more than cortical.
• Frontal cortex only: COx samples ~2 to 3 cm depth in frontal cortex; says nothing about deep brain, posterior fossa, or contralateral hemisphere.
• Mixed venous-arterial compartment: rSO2 is ~75% venous, ~25% arterial; not pure tissue oxygen tension.
• Pad failure: poor contact, signal drift, drop-out; check quality continuously.
• Persistent low signal quality: do not interpret COx; consider alternative autoregulation index.
• Discordance with PRx interpreted as error: sepsis-related microvascular shunting is a real physiology, not an artefact.
• Single-snapshot COx: trend over hours is the signal.
• Pediatric / neonatal COx normative ranges: still emerging; adult thresholds may not directly apply.
• SafeBoosC ≠ COx targeting: SafeBoosC targets rSO2 within a band, not COx; the COx framework is the underlying physiology but the bedside protocol is rSO2-based.
• MAPopt validity in low MAP variation: the U-curve fit is unreliable when MAP has not swung across enough of its range; needs hours of natural variation.

14. Combine with…#

• NIRS: the parent modality; COx is one of its derived indices.
• PRx: the invasive autoregulation sibling.
• Mx: the TCD-based non-invasive autoregulation sibling.
• CPP: the upstream hemodynamic variable.
• CPPopt: the dedicated CPPopt page; COx-MAPopt is the NIRS-based variant.
• Foundations: autoregulation: the underlying physiology.
• Advanced NIRS: DCS and TR-NIRS as emerging absolute CBF tools.

15. Evidence summary#

16. Recent literature (2022 to 2025)#

• Plomgaard 2024 SafeBoosC-III primary results: did not show mortality benefit at 36 weeks for rSO2-guided preterm care vs standard care; framework remains established in many units. 3
• Hansen 2023 SafeBoosC-III 2-year follow-up: confirmed no significant outcome difference at 2 years. 5
• Naim 2023 pediatric post-arrest brain injury: NIRS / COx as part of multimodal post-arrest framework. 6
• Helbok 2024 pediatric MMM: COx as an accessible pediatric autoregulation index in resource-stratified centres. 26
• Figaji 2025 pediatric MMM consensus: positions COx alongside PRx and Mx in the pediatric MNM autoregulation triad. 25
• Continued NIRS device evolution: time-resolved NIRS, frequency-domain NIRS, and DCS expand the NIRS family with absolute CBF and tissue oxygenation indices. 36

17. Self-check#