Will the Compensatory Reserve Index be useful for anaesthetists?

Moulton SL et al (2017) “Validation of a noninvasive monitor to continuously trendindividual responses to hypovolaemia” J Trauma Acute Care Surg 83(1)s1:S104-11

“‘Til the roof comes off, till the lights go out
‘Til my legs give out, can’t shut my mouth.” – Nate Dogg

Short Summary

This article reports on a novel monitor for cardiovascular reserve, the Compensatory Reserve Index (CRI). The CRI was developed to assist with diagnosis of concealed blood-loss in the pre-hospital trauma setting, and predicts a patient’s cardiovascular reserve using artificial intelligence methods to compare the plethysmograph waveform to a reference library of waveforms at progressive simulated hypovolaemia. The study population were venesected of 20% estimated blood volume, aiming to characterise rate of change and nadir of CRI in patients with and without hypotension after this blood loss.

We felt that the study probably did show that CRI correlates with degree of cardiovascular reserve with progressive venesection, but that it was difficult to make clear comparisons of the study subjects because of the way the study was reported. There was incomplete demographic data, and pre- and post-venesection CRI data was reported as a single undifferentiated group, rather than separating the study population into hypotension and no-hypotension groups for the entire report.

This study is challenging to assess, as due to the novel nature of CRI, there is no clear gold standard against which to compare it. Concerns were raised that performing a comparison to a library of plethysmograph waveforms is fundamentally only as good as the waveform library, and the demographics of the subjects used to populate the library have not been clearly described – other than being healthy adults. No work has been done to validate CRI in the extremes of age, in pregnancy or in comorbid patients – particularly those with cardiovascular disease; on those with non-hypovolaemic causes of shock (cardiogenic / obstructive / distributive / dissociative); or in the setting of vasoactive drugs. Of note, later studies have begun to assess CRI in paediatric sepsis, and a baboon venesection validation study has been performed.

Despite these concerns, the CRI, shows potential as a non-invasive measure of cardiovascular reserve in the perioperative patient, and in the pre- and post-operative setting. We would be interested to compare the CRI with an invasive cardiac output monitor in the perioperative setting.

Summary by Dr Ged Manning

Detailed Summary


A group in the University of Colorado (US Army funded) started out with the concept of “Concealed blood loss is a major/difficult to diagnose cause of pre-hospital death in trauma. How can we better identify/quantify it?”. The intended use was assessment of prehospital trauma patients to detect life-threatening concealed blood loss. After investigating commonly used metrics (RR/HR/SPB/DBP/Pulse Pressure) and finding they weren’t particularly helpful in young fit subjects, they decided to try plethysmograph pulse contour analysis.

They devised the “Compensatory Reserve Index” (CRI) – using artificial intelligence techniques to analyse plethysmograph waveforms and compare them to a database of known plethysmograph waveforms of individuals who had progressive hypovolaemia. They created their database by subjecting 184 healthy 18-55 year-olds to progressive negative lower body pressure, causing lower body blood pooling as a simulation for hypovolaemia[1]. The initial paper describes the baseline patients as “healthy, non-smoking normotensive males or females, with ages ranging from 18 years to 55 years”, with no other demographics or more detailed analysis. The subjects were subjected to progressive reduced lower body pressure until systolic blood pressure < 80 mm Hg with or without bradycardia, and the CRI was defined as a decimal on a scale where 1 = maximal reserve to blood loss and 0 = onset of systolic blood pressure < 80 mm Hg with or without bradycardia, indicating complete depletion of reserve (figure 1). It is not clear to me if baseline BP, or the presence of symptoms of hypovolaemic shock, were considered. The CRI is calculated based on the 30 previous pulse waveforms, and is updated after every heartbeat.

Figure 1: Concept graph of CRI number against time/progressive blood loss, from health to decompensation/death.

Prior to the study discussed in this journal club, the CRI has been compared to HR/BP/Sats/Stroke index/Base Deficit in lower body negative pressure in other patient groups[2] [3] [4] [5], 1 unit blood donation[6] [7], ~1.2 litre blood loss[8], in baboons to 25% EBL[9]. It had also been trialled in injured adults in a trauma centre[10]. The research papers investigating CRI all have authors in common. This work has shown that CRI appears to correlate in a linear way to actual blood loss, and have a greater sensitivity/specificity for predicting time to systolic blood pressure < 80 mm Hg or symptomatic hypovolaemia than conventional metrics. CRI of 0.6-1 has been defined as “normal volume reserve”, of 0-0.3 as “entirely depleted volume reserve”, and 0.3-0.6 as in-between.

The study considered in this journal club intended to assess the CRI response to 20% estimated blood loss by sequential venesection, hypothesising that the nadir CRI values would be lower in subjects experiencing symptoms of hypovolaemia or SBP < 80mmHg, and that subjects who were less tolerant of blood loss would experience more rapid reduction in CRI.


A prospective, single-centre validation study was performed.

42 healthy volunteers were selected, with exclusion criteria BMI > 30, smokers, medication (other than OCP) and pregnancy. Subjects were instructed to sleep normally and avoid exercise, caffeine & medication for >24 hours.

Subjects had ECG & “continuous” NIBP monitoring.

Each subject had 4 CRI monitors attached – three switched on throughout, one which was turned off between measurements, to assess the impact of any device “memory” as blood loss progressed.

A large-bore ACF cannula was placed.

Estimated total blood volume was defined as male 75 ml/kg; female 65 ml/kg.

20% estimated blood volume (max 1.3 litres male; 1 litre male) was removed in 333ml aliquots.

There was a 3 minute pause between aliquots for data collection.

Blood was returned on hypotension (SBP <80 / >30% below baseline), symptoms of hypovolaemia, or target blood loss achieved.

Data collected wasn’t clearly described, but appears to be a static level of CRI after each 333ml blood removal, as well as the entire CRI trace for the experiment.

Blinding wasn’t attempted.

Power wasn’t discussed.

Statistical tests were not explicitly pre-specified. Comparisons were made between subjects who experienced hypotension (SBP <80 / >30% below baseline) or “symptoms” of hypovolaemia; and subjects who were not compromised after 20% EBL. Two-sample T-Tests were used to compare CRI values between groups, and a linear mixed model with CRI as outcome and volume loss as predictor.


42 healthy volunteers (24M:18F)
                   |- 3 excluded (2 “data corruption”, 1 unable to draw blood)
     20% blood removed
          |                            |
    32 not                      7 symptomatic
symptomatic                  4M:3F

Baseline characteristics:

  • 24 M : 18 F
  • 19-36 years
  • 1.55 – 2.01 meters
  • 52.6 – 101.6 kg
  • BMI 20-29.4

No comparison of characteristics between symptomatic/asymptomatic groups – not reported, so unable to apply Carlisle Method.

No additional baseline characteristics/demographics given

Primary Outcomes:

Did they have two separate populations?

The subjects with hypotension (SBP <80 / >30% below baseline) +/- symptoms had a lower total blood loss than the patients who did not have hypotension or symptoms (figure 2). No statistical comparison was reported of the total blood loss, but visually there appears to be a difference between the two groups.

Figure 2: Total blood loss for subjects hypotension (SBP <80 / >30% below baseline) +/-  symptoms; and those without.

Hypothesis: CRI values will be lower in subjects experiencing symptoms of haemodynamic decompensation

CRI values were reported graphically and as mean/95% confidence interval. CRI prior to blood loss in all subjects appears to be higher than following blood loss, and then appeared to return to near-normal levels. The CRI values associated with hypotension/symptoms were in the region of 0-0.3, in keeping with previous work.

Importantly, although the CRI after blood loss has been split into two groups (those with and without hypotension), the CRI values at the start and end of the study have not, making it impossible to demonstrate a clear progression of events between the two groups. No statistical test were described to compare the two groups.

Figure 3: CRI before (all subjects); at maximal blood loss (separately for subjects without and with hypotension/symptoms); and after blood replacement (all subjects).

Hypothesis: Subjects with lower tolerance to blood loss will have a more rapid fall in CRI

The asymptomatic group had a ∆CRI = 0.28 / Litre blood loss [95% CI 0.24-0.33], and the symptomatic group had a ∆CRI = 0.85 / Litre blood loss [95% CI 0.67-1.0] (two-sample t-test p<0.01)

Secondary Outcomes

Correlation between CRI and volume of blood loss = 0.92 (95%CI = 0.87-0.95)

“No difference” between the different CRI units (not further expanded upon)

“No difference” between the CRI units left on continuously, and those turned off between measurements (not further expanded upon)

CRI trajectories in symptomatic patients

CRI trajectories for the period around symptoms and blood return were reported for the symptomatic patients (figure 4)

Figure 4: CRI values against time, for the symptomatic patients, around the time of symptoms


Paper: “CRI effectively trends the proportion of additional volume loss an individual can tolerate before CV collapse”

I agree – CRI <0.3 appears to predict hypotension+/- symptoms of hypovolaemia in progressive venesection to 20% EBL in this patient group.


The entire concept of the CRI is to compare a 30-beat stretch of the plethysmograph trace with a library of plethysmograph traces in progressive simulated hypovolaemia, and to give a decimal description of how far between baseline and SBP<80mmHg a waveform analysis “AI” algorithm assesses the patient to be. Fundamentally, this depends on the generalisability of the library population – and the library population don’t seem to have been well described in terms of age/sex/ethnicity/nationality/demographics.

I feel that this paper doesn’t do justice to the study it reports, as it sets out to separate the population into two groups (those with and without hypotension +/- symptoms after 20% blood loss), but then doesn’t clearly/explicitly compare these two groups as two separate groups. Despite this, I feel that they have demonstrated that CRI is lower in subjects who are currently experiencing hypotension/symptoms of blood loss, and that the CRI falls more rapidly in subjects who tolerate 20% blood loss less well.

There is partial description of methodology and results to compare multiple CRI units, and CRI units which are continuously turned on compared to being intermittently turned on. This confuses the overall message of the paper, without providing enough data to justify the stated conclusions.

Very few statistical analyses are reported. Much of the analysis is simply descriptive rather than comparative, but concerningly, data which is clearly not normally distributed in graphical form (CRI absolute value) is described using mean/standard deviation/95% confidence interval, rather than median / interquartile range / 95% confidence interval. No comparative statistical tests appropriate for assessment of non-normally distributed data are mentioned in the methods section, although no comparative statistics are reported.

I do not feel that I can assess the external validity of this study from this paper, given the lack of baseline demographical information about the patient group. The Carlisle method could not be applied to this paper for this reason. As neither this study population, nor the original population used to produce the CRI waveform library, have been well-defined, it is impossible to assess if the two populations are similar.

Specific to potential applications in perioperative care, the utility of CRI is unclear due to the inadequate description of the population used to create the baseline library of plethysmograph traces, and the lack of data on broader patient groups (extremes of age, comorbidities and in cardiovascular disease), on non-hypovolaemic causes of shock (cardiogenic / obstructive / distributive / dissociative) and on interactions with vasoactive drugs.

Despite the limitations in reported data, the reported results of CRI show promise. Unpublished CRI/time graphs for a patient experiencing induction of general anaesthesia (figure 5) and a major obstetric haemorrhage (figure 6) correlate well with my clinical impression of the patieants volume reserve, which I know to be frequently less clearly shown by traditional monitoring. CRI has been shown to predict perforation in paediatric appendicitis[11] (figure 7), and to improve with clinical improvement in dengue shock syndrome[12] (figure 8), suggesting a possible generalisability in sepsis.

A non-invasive assessment of volume status is an attractive prospect, and plethysmograph waveform analysis carries enough similarity to arterial blood pressure waveform analysis (LIDCO etc.) to be a plausible approach. I hope that the attractive proposition does not mask the complexities and potential pitfalls of AI waveform analysis.

Figure 5: CRI against time during induction of general anaesthesia (details of patient unclear)
Figure 6: CRI against time during an epidural top-up LSCS with PPH (details of patient unclear)
Figure 7: CRI in perforated and non-perforated appendicitis in children.
Figure 8: CRI in paediatric dengue shock syndrome

Summary by Dr Ged Manning


[1] Moulton SL et al. (2013) “Running on empty? The compensatory reserve index.” J Trauma Acute Surg 75(6):1053-9 https://www.ncbi.nlm.nih.gov/pubmed/24256681

[2] Van Sickle C et al. (2013) ” A Sensitive Shock Index for Real-Time Patient Assessment During Simulated Hemorrhage” Aviation, Space & Environmental Medicine 84(9):907-12https://www.ingentaconnect.com/content/asma/asem/2013/00000084/00000009/art00002

[3] Convertino VA et al. (2013) “Estimation of individual-specific progression to impending cardiovascular instability using arterial waveforms” J Appl Physiol 115(8):1196-1202 https://www.physiology.org/doi/full/10.1152/japplphysiol.00668.2013

[4] Janak JC et al. (2015) “Predictors of the Onset of Hemodynamic Decompensation During Progressive Central Hypovolemia: Comparison of the Peripheral Perfusion Index, Pulse Pressure Variability, and Compensatory Reserve Index” Shock 44(6):548-53 https://journals.lww.com/shockjournal/Fulltext/2015/12000/Predictors_of_the_Onset_of_Hemodynamic.6.aspx

[5] Howard JT et al. (2016) “Specificity of Compensatory Reserve and Tissue Oxygenation as Early Predictors of Tolerance to Progressive Reductions in Central Blood Volume” Shock 43(3S):68-73 https://journals.lww.com/shockjournal/Fulltext/2016/09001/Specificity_of_Compensatory_Reserve_and_Tissue.10.aspx

[6] Roy N et al. (2014) “The Value of Noninvasive Measurement of the Compensatory Reserve Index in Monitoring and Triage of Patients Experiencing Minimal Blood Loss” Shock 42(2):93-8 https://journals.lww.com/shockjournal/Fulltext/2014/08000/The_Value_of_Noninvasive_Measurement_of_the.3.aspx

[7] Stewart CL et al. (2014) “Detection of low-volume blood loss: Compensatory reserve versus traditional vital signs” J Traum Acut Surg Care 77(6):892-8 https://journals.lww.com/jtrauma/Fulltext/2014/12000/Detection_of_low_volume_blood_loss___Compensatory.14.aspx

[8] Convertino VA et al. (2015) “Individual-Specific, Beat-to-beat Trending of Significant Human Blood Loss: The Compensatory Reserve” Shock 44:27-32 https://journals.lww.com/shockjournal/Fulltext/2015/08001/Individual_Specific,_Beat_to_beat_Trending_of.5.aspx

[9] Carneb HL et al. (2016) “Comparison of compensatory reserve during lower-body negative pressure and hemorrhage in nonhuman primates” Fluid and Electrolyte Haemostasis 310(11):R1154-9 https://www.physiology.org/doi/full/10.1152/ajpregu.00304.2015

[10] Stewart CL et al. (2016) “The Compensatory Reserve Index Following Injury: Results of a Prospective Clinical Trial” Shock 43(3S):61-7 https://journals.lww.com/shockjournal/Fulltext/2016/09001/The_Compensatory_Reserve_Index_Following_Injury__.9.aspx

[11] Choi YM et al. (2017) “Noninvasive monitoring of physiologic compromise in acute appendicitis: New insight into an old disease” J Ped Surg 53(2):241-6  https://www.jpedsurg.org/article/S0022-3468(17)30731-5/abstract

[12] Moulton SL et al (2016) “State-of-the-art monitoring in treatment of dengue shock syndrome: a case series” J Med Case Rep 10(1):233 https://www.ncbi.nlm.nih.gov/pubmed/27553703

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s