Option 1: CPP-PRx error bar
As described in the section ‘ICM+’ there are many options to explore or customize the optimal cerebral perfusion pressure (CPPopt) algorithm in the ICM+ program. At present this function is mainly applied using the parameters intracranial pressure (ICP) and arterial blood pressure (ABP) in TBI and SAH patients. In the current section the ‘CPP-PRx error bar’ where the U shaped curve can be fitted by an automated mathematical algorithm and the CPPopt value automatically displayed will be described.
Plotting the PRx values against CPP over a certain time period (for example 4-6 hours) represents the backbone of this methodology. The error bar chart ‘configuration’ panel shows the different options available at the moment (figure 1 and table 1). Below a simplified and hopefully practical explanation of the options is provided. For more details we refer to CPPopt publications or contact the authors.
On the left side of the chart configuration panel (figure 1) the signals can be selected for the graph with their lower and upper limits. For plotting the classical ‘CPPopt’ curve the signals CPP and PRx are selected.
With selecting 40 and 120 mmHg as the lower and upper limits and dividing this range by for example 16 (subsets) CPP bins of 5 mmHg are created to represent the mean PRx value (± S.E.) in the error bar plot. For different patient populations different ranges and subsets might be needed (to illustrate: neonates need a different CPP range selection). For the error bar display (series) additional selections can be made (different signals, max/min, error bar style, colour and line size).
Automated CPPopt curve fitting can be turned on using ‘Optimal curve fit enable’ check box on the right side of the configuration panel (figure 1). Below that checkbox options for curve fitting and presentation can be selected. In principle, an algorithm described in details in Aries et al. fits a second order polynomial curve to the means of Fisher transformed and binned PRx values and finds the minimum y-value (PRx), termed PRxopt [2]. The corresponding x-value (CPP) is automatically displayed as the CPPopt value.
In table 1 the different options (heuristics) to influence or restrict the curve fitting process are explained in more detail and can help you to choose your final settings. We will provide the most common settings used so far (default settings). Please keep in mind that the CPPopt research line is very dynamic and frequently new options and features become available 🙂
Figure 2. The CPP-PRx error bar without fitted curve.
Figure 3. The CPP-PRx error bar with a fitted curve.
Figure 4. The CPP-PRx error bar with the additional ‘Optimal CPP range’ which is the CPP range marked in between the red boxes (PRx threshold of 0.25).
Additional explanation of recent configuration panel options
0 Weigh the fit by the error size
If set this option will force the curve fitting to be done with weighting by the size of the error bars. The larger the error bar the less influence that mean value will have on the curve fit. By default all the mean values are treated the same.
1Reject if curve is outside the range
If this function is used (enabled by the checkbox), the fitted curve (evaluated within the range of CPP value used for the fit) that is entirely out of the set PRx range will be rejected. If set appropriately (e.g. to a range of -0.3 to 0.6) this option ensures that in the conditions of entirely impaired or entirely intact autoregulation (within the probed CPP range) there would be no CPPopt recommendation given.
2 Optimal autoregulation CPP range
The fitted CPP-PRx curve can extrapolated at both sides to include the full range of plausible CPP values (from 40 to 120 mm Hg) to obtain the CPP values at which the curve crosses the selected threshold PRx value for impaired pressure reactivity (e.g. PRx= +0.25). The values of CPP at these two points of intersection are labelled as the ‘lower limit of reactivity’ (LLR) and ‘upper limit of reactivity’ (ULR), respectively (figure 4). If a curve is entirely above the PRx threshold, no intersection could be calculated and thus the LLR and ULR will set to equal the CPPopt (i.e. no other values can be retrieved from the curve). The value of 0.25 as a threshold for PRx was chosen as it has been identified as a critical threshold for determining fatal outcome in a previous study of severe TBI patients in our TBI cohort.[3]
In addition, as a protection against extreme, unphysiological values for the limits of reactivity that could be returned from the curve fit LLR and ULR returned are limited to the total range of valid CPP values set for the chart (i.e. commonly 40 to 120 mmHg). The same approach is applied to situations where the fitted CPP-PRx yields a monotonically ascending or descending curve with no inflection point.
Table 1. CPP-PRx error bar settings
| SETTING | DESCRIPTION | DEFAULT setting (adults) |
|---|---|---|
| Variable | X (binning) variable, in the current context should be set to Cerebral Perfusion Pressure (CPP) variable with calculation window the same as the Y variable (ie 5min for PRx) | CPP5min |
| Low Limit | Minimum x-value | 40 mmHg |
| High Limit | Maximum x-value | 120 mmHg |
| Subsets | The number of subsets for the X variable is chosen so that each bin covers a certain CPP range. | 16 subsets (CPP bins of 5 mmHg) |
| Variable | Y (bins descriptor) variable, in the current context should be set to pressure reactivity (PRx) variable | PRx |
| Max | Maximum y-value | Automatic |
| Min | Mininimum y-value | Automatic |
| Bar Style | Display setting | Optional |
| Exclude if less [%] | Relative minimum number of data points that has to be included in the error bar. If this specific error bar contains less data than the set percentage, no error bar will be given (and not used for curve fitting). | 2% |
| Error line size | Display setting | Optional |
| Optimal curve fit Enable | Enable the curve fitting | Checked |
| Concave shapes | If checked, it will be attempted to look for a maximum in the curve rather than a minimum value (convex shape). | Not used (so far) |
| Min. data incl. [%] | Relative minimum number of data points (in all the bins included in the curve fit) that has to be present for the curve fit to return a valid value. | 50% |
| Min. fit span | The fitted curve must span a certain PRx range, otherwise the curve is rejected. | 0.10 or 0.20 |
| Need not include ‘best’ value | ‘Best’ CPP refers to the CPP bin with the lowest PRx value in the calculation period. If set, this option will only allow curve fitting when the curve covers the ‘best’ CPP bin. | Optional |
| Weigh the fit by the error size | Use the size of the error bars as weights for curve fitting. | Optional |
| Reject if curve is outside the range | If checked, the curve that is entirely out of the set range will be rejected. Insert the PRx range. | Optional |
| Show optimal (CPP) range | If checked, the chart will highlight the areas and CPP range outside the lower and upper limit of pressure reactivity (LLR/ULR). A PRx threshold should be chosen. | Optional |
| Optimal (CPP) range threshold | The PRx threshold inserted, to define the ‘optimal’ CPP or ‘safe’ CPP range. | 0.25 |
| Curve fit line size | Display setting | Optional |
References
[1]. Aries MJ, Wesselink R, Elting JW, Donnelly J, Czosnyka M, Ercole A, Maurits NM, Smielewski P. Enhanced Visualization of Optimal Cerebral Perfusion Pressure Over Time to Support Clinical Decision Making. Crit Care Med. 2016 Oct;44(10):e996-9.
[2]. Aries MJ, Czosnyka M, Budohoski KP, Steiner LA, Lavinio A, Kolias AG, Hutchinson PJ, Brady KM, Menon DK, Pickard JD, Smielewski P. Continuous determination of optimal cerebral perfusion pressure in traumatic brain injury. Crit Care Med. 2012 Aug;40(8):2456-63.
[3]. Sorrentino E, Diedler J, Kasprowicz M, Budohoski KP, Haubrich C, Smielewski P, Outtrim JG, Manktelow A, Hutchinson PJ, Pickard JD, Menon DK, Czosnyka M. Critical thresholds for cerebrovascular reactivity after traumatic brain injury. Neurocrit Care. 2012 Apr;16(2):258-66.