Ethics approval
The study was approved by the Regional Committee for Medical and Health Research Ethics, Southern Norway (REC ID: 2012/1155). All participants gave written informed consent. The study followed the Declaration of Helsinki and was conducted according to Good Clinical Practice.
Patient population
Healthy, non-smoking, normotensive women with singleton pregnancies scheduled for elective caesarean delivery under spinal anaesthesia were asked to participate. Participants were recruited in collaboration with "The Placenta Project" (REC ID: 2011/2419) and are a subset of the study population of that project. As part of the study protocol for "The Placenta Project", all participants received an intra-arterial line used for blood sampling at delivery. The protocol for the placenta study is published previously[4]. Exclusion criteria were considerable pre-existing morbidity, pregnancy complications, contractions prior to scheduled C-section, and also prior Raynaud phenomena, as this is not compatible with use of the ccNexFin monitor. Two women with hypothyroidism, each supplemented with a low dose of L-thyroxine (50 and 75 mg daily, respectively) and one woman with mild asthma and occasional use of salbutamol (not taken in the days prior to participation in the study), were included. The inclusion period was from May 2013 until January 2014. Demographic properties are shown in table 1.
|
Mean (SD)
|
Range
|
Age (yr.mo)
|
35.10 (3.2)
|
29.6 – 42.7
|
Height (cm)
|
167 (5.0)
|
160 – 180
|
Weight before pregnancy (kg)
|
64.0 (10.7)
|
50 – 91
|
Weight at delivery (kg)
|
78.9 (12.3)
|
60 – 105
|
Length of pregnancy (days)
|
275 (6.5)
|
260 – 292
|
Table 1: demographic properties of the participants.
Monitoring devices
The LiDCOplus monitor is in routine clinical use and has documented accuracy and trending abilities[5]. It provides information about circulatory changes from heartbeat to heartbeat, and is used when advanced monitoring is indicated, such as during major surgery, or during interventions on patients with circulatory disorders. Mathematical analysis of the intra-arterial pressure curve by pulse power analysis is done using the built-in software PulseCO. It estimates many aspects of the circulation and can be used with or without lithium dilution calibration (LiDCO). In this study, we used calibrated CO, aiming at optimal accuracy. The technology has been used in studies of healthy women and pregnant women with heart disease [6], and is the standard method for perioperative monitoring of pre-eclamptic women at our institution. In the comparison of CO methods it serves as the reference.
The ccNexfin monitor is based on the principle of the unloaded vascular wall [7], the Physiocal criteria [8], and a generalized waveform filter to reconstruct brachial pressure from finger pressure [9]. An inflatable cuff is placed around one of the three middle fingers of either hand. An integrated plethysmograph measures the volume of blood under the cuff using an infrared light source and a photosensor. The monitor initially determines a set point for the finger cuff pressure, where most of the venous blood is displaced, and the arterial diameter is reduced to no more than 50% of the expanded diameter. The set point is intermittently calibrated according to the Physiocal criteria, to account for changes in the vascular state of the finger. The cuff pressure is continuously adjusted to counter the varying intra-arterial blood pressure, keeping the signal from the photosensor, and consequently the blood volume and arterial diameter under the cuff, constant. This way, the artery wall is said to be unloaded, transmural pressure is zero, and the pressure in the cuff thus represents the intra-arterial pressure. The measured finger blood pressure is transformed to reflect brachial blood pressure. The CO calculations in the ccNexFin are based on pulse contour analysis of the derived arterial pressure curve. The monitor is designed to work without external calibration. Our research group has been involved in several projects using finger plethysmographic monitor technology, including the largest population-based study ever utilizing this technology, The Tromsø Study [10].
Conduct of the study
Design: A prospective observational study. Monitoring: A 20G BD arterial cannula (Becton Dickinson Infusion Therapy Systems Inc, Utah, USA) was placed in the radial artery after skin infiltration with lidocaine (5–10 mg). It was connected to a Siemens Dräger Infinity Gamma XL hemodynamic monitor (Drägerwerk AG & Co. KgaA, Lübeck, Germany) via a Codan X-trans pressure transducer (CODAN pvb Critical Care GmbH, Forstinning, Germany), and the signal calibrated according to standard departmental procedures. Peripheral IV catheters were placed on both arms.
Intra-arterial blood pressure data was passed through to the LiDCOplus monitor. Heart rate (HRinv), systolic arterial pressure (SAPinv) and mean arterial pressure (MAPinv) was recorded at a rate of one sample per heartbeat. COLiDCO, estimated by PulseCO, was also recorded. A single point calibration of CO was performed. The ccNexFin monitor was applied to one of the three middle fingers on the same arm as the intra-arterial cannula. Corresponding variables (HRnex, SAPnex, MAPnex and COnex) were recorded.
While sitting on the operation table, the subjects received spinal anaesthesia with bupivacaine 10 mg and fentanyl 20 µg, using a 27G pencil point needle. Co-loading with IV NaCl 0.9% 1000 mL was started. The parturients were then placed in the supine position with left lateral tilt, using a wedge under the right hip. Immediately after injection of the drugs, an IV bolus of phenylephrine 25-50 µg was given, followed by an infusion started at 0.25 µg/kg/min and titrated according to invasive blood pressure, aiming for a stable SAPinv > 90 mmHg.
Data recording
In order to acquire synchronous sampling from both monitors, measurements were sampled in real time by the same computer. Samples were acquired through the RS232 port of the LiDCO monitor and the analogue output from the ccNexfin monitor using a data acquisition card and software from National Instruments. This setup was evaluated for electrical safety and approved by the appointed committee at Oslo University Hospital.
Time-stamped data for inter-beat interval (IBI), SAP, MAP, and CO from both monitoring devices was recorded to a single dataset per subject, one sample per heartbeat, using software developed in-house in National Instruments LabVIEW®. Events were marked in real time and saved to file using the same software.
Due to subject movement following spinal anaesthesia, placement of a hip wedge, and adjustment of the arterial pressure transducer and the ccNexFin heart reference system, we considered data from the first two minutes after spinal anaesthesia unreliable. The arterial line was used for blood sampling just prior to delivery, causing a pause in our registration, and following delivery, there was again more subject movement causing unreliable data. It is also the experience of our group that the LiDCO needs recalibration after delivery [11]. We included the data between 2 and 12 minutes after spinal anaesthesia for our calculations.
Statistics
Due to differences in processing time between the two monitors, the LiDCO samples were ahead of the ccNexfin samples, on average by around two heartbeats, although they were sampled synchronously from the outputs of the monitors. For each session, this difference was adjusted by calculating the lag by means of a cross-covariance analysis of the IBI time-series, which are assumed to be equal between monitors, and then shifting the ccNexfin recording ahead by the calculated lag in order to align the recordings for comparison at equal beats. This was done in Matlab R2014b (Mathworks, Nantick, Massachusetts, UAS.)
Artefacts were reduced using a previously published method for detecting and removing outliers in continuous blood pressure and cardiac output recordings [12]. Data points for statistical analysis were constructed with one-minute intervals, by averaging data over the first 10 seconds of each minute.
Method comparison statistics were done using Matlab and Stata v15 (Statacorp LLC, College Station, Texas, USA). We used the method first described by Bland and Altman to investigate the agreement of the ccNexFin monitor with invasive blood pressure and LiDCO cardiac output measurement [13,14]. Early versions of this method did not sufficiently consider structure in the data, and could produce too narrow limits of agreement, and too narrow confidence intervals, with repeated measurements per subject. We calculated limits of agreement based on the repeated observations method as described by Zou [15]. Confidence intervals for the limits of agreement were calculated using the MOVER algorithm [15]. This method is preferred when the true value varies, when there is a different number of measurements from each subject, and the between-subject variance is large with respect to the within-subject variance. The MOVER method also allows the construction of asymmetric CIs. Diagnostic plots suggested by Bland and Altman were inspected to check for underlying assumptions.
Polar plots were used to assess trending abilities for both blood pressure and CO [16]. As suggested by Critchley, the smallest changes were considered most likely to represent noise, and excluded from the polar plot analysis. Data points with an average change of more than 5 mmHg (for BP) or 0.5 L/min (for CO) from the previous measurement were included.