This study was a secondary explorative analysis of data from another randomized clinical trial analyzing serum fentanyl levels in parturients randomized into epidural analgesia with and without the addition of adrenaline 2 µg.ml− 1 to the solution.[10] The study was approved by the regional ethics committee (REK Sør-Øst, ID number 2012/32, approved March 2012) and the Norwegian medicines agency, registered at clinicaltrials.gov (NCT00685672), and conducted according to Good Clinical Practice guidelines. The reporting in this manuscript adheres to the CONSORT guidelines.[11] All participants gave oral and written informed consent. The study design was a randomized controlled trial with two parallel groups. The parturients, the investigators, all personnel treating the participants and assessing the outcomes were blinded to the intervention.
Inclusion criteria were American Society of Anesthesiologists (ASA) class I and II adult (> 18 years) singleton nulliparous women in active labor requesting epidural analgesia. Exclusion criteria were pre-gestational body mass index (BMI) > 35 kg.m− 2, height < 155 cm, reduced communication skills in Norwegian or English, known hypersensitivity to medications used in the solution or other contraindications to epidural catheter placement. Women with preeclampsia was classified as ASA class 3 and thus excluded. All participants were recruited at the birth clinic at Akershus University Hospital, Loerenskog, Norway, which has approximately 5000 deliveries annually. Patients were included from June 2014 to September 2015.
A multi orifice epidural catheter (PERIFIX, B-Braun, Melsungen, Germany) was inserted 5 cm in the epidural space at L1-2 or L2-3 by an 18 gauge Tuohy needle using the loss of resistance technique with saline, with the patient in the sitting position. The skin was anesthetized using lidocaine 10 mg.ml− 1 without adrenaline.
The patients were randomized to receive either an epidural solution of bupivacaine 1 mg.ml− 1 and fentanyl 2 µg.ml− 1 (control group) or bupivacaine 1 mg.ml− 1, fentanyl 2 µg.ml− 1 and adrenaline 2 µg.ml− 1 (adrenaline group ). The blinded test drug solution bags were manufactured by the Hospital Pharmacy at Oslo University Hospital – Rikshospitalet according to the randomization list. The randomization list was created by a researcher who did not take active part in the study (see acknowledgements), using a list of random numbers.[12] Test drug bags were marked with general information of the study, the constituents – including information about containing adrenaline or placebo – and study number. After epidural catheter placement, 5 mL of the solution was injected as a test-dose. If no signs of vascular or intrathecal catheter placement were found, an additional 5 ml of the solution was injected, and a continuous infusion of 5 mL.h− 1 of the solution was started using an infusion pump (CADD-Legacy PCA®, Smith medical, St Paul, MN, USA). The participants had the possibility of patient controlled epidural boluses (PCEA) of 5 ml, with a lock-out time of 30 minutes, and were instructed to use this option if pain relief was inadequate.
Before epidural catheter placement, blood pressure was measured once at the arm using an automated oscillometric blood pressure monitor (ProCare 100, GE Medical Systems Information Technologies, Inc., Milwaukee, Wisconsin) between contractions. In addition, the patient was connected to a Nexfin CC® monitor (BMEYE B.B., Amsterdam, The Netherlands). The Nexfin CC is a non-invasive continuous blood pressure and cardiac output monitor using the volume clamp method for measuring blood pressure[13, 14], transforming it to a brachial blood pressure waveform. The cardiac output is estimated by the pulse contour method.[15] The Nexfin has been validated for blood pressure measurements in pregnant women,[16] and for cardiac output in cardiac surgical patients.[17, 18]
The monitor was calibrated to the patient’s height, pre-delivery weight, age and sex. The patients were monitored for 60 minutes after epidural activation. A marker for ‘time = 0’ was created in the Nexfin CC monitor when the second starting bolus was given.
Fluids and/or vasopressors were given at clinical indications (fetal or maternal), no mandatory pre- or co-loading of fluids were given according to general departmental procedure.
All patients were monitored using cardiotocography, with use of ST-segment analysis at the discretion of the attending midwife or obstetrician according to standard procedure.
Data extraction and preparation
Data was stored in the Nexfin device as proprietary .csd, .xml and .idx files. The data were extracted using the FrameInspector ® software (v 2.3.0.2, BMEYE B.B, Amsterdam, The Netherlands), and beat-to-beat data were converted to Microsoft Office Excel 2010 ® (Microsoft, Redmond, WA, USA) format, and then imported to MATLAB® version R2015b (MathWorks, Natick, MA). In MATLAB, the data were first cleared of artifact readings using an algorithm previously published[19], which uses the deviation from the median of surrounding measurements in combination with appropriately selected threshold values for each parameter (cardiac output: 5, heart rate: 25, Systolic blood pressure: 50, MAP: 50, Diastolic blood pressure:50, SVR: 400). Tracings for all patients were manually inspected to ensure correctness of artifact clearance. Secondly, the beat-to-beat measurements were transformed to median values with a frame of 50 measurements and 50% overlap between frames to acquire a smoother curve. This step was also manually inspected to ensure that the median values represented the original data. Finally, the peak at each contraction was found using the findpeaks() function in MATLAB, using a minimum criterion for the peak prominence (how much the peak stands out due to its intrinsic height and its location relative to other peaks), adjusted manually for each recording (2–10 mmHg for blood pressure peaks and 0.1–0.7 L.min− 1 for cardiac output). These peak values were used as the primary outcome and as dependent variables in the final analysis. Figure 2 gives an example of data from a representative case.
Statistical analysis
The primary outcomes in this secondary analysis were changes in cardiac output and systolic blood pressure related to contractions. The hypothesis was that the change over time in these hemodynamic outcomes following epidural activation is different between the two study groups (with and without adrenaline). The temporal development of cardiac output and systolic blood pressure was first inspected visually within the 30 minutes window following epidural activation by plotting these variables against time. Because no particular curve or phase transition could be identified in the data, a linear approach was assumed for statistical analysis. A linear mixed-effects (LME) model was employed for testing this hypothesis based on the interaction effect between time (the time at the identified peaks of the respective variable, from epidural activation to 30 minutes later) and adrenaline (binary variable for study group with or without adrenaline in the bolus). With systolic blood pressure and cardiac output as the dependent variable in different tests, time, adrenaline and the time-adrenaline interaction were used as fixed effects, with subject as a random effect having a random intercept and slope (for the effect of time). The p-value and confidence intervals of the time-adrenaline interaction effect were used to draw inferences on the effect of adrenaline on changes in hemodynamics after epidural activation. The 30 minutes timeframe was chosen to capture any potential hemodynamic changes due to both the sympathetic block caused by the epidural, and the potential effect of adrenaline entering the systemic circulation, primarily at the initial boluses. Previous studies have shown the time to maximum skin temperature change (due to sympathetic block) in the lower extremities is approximately 15 minutes when using epidural analgesia[20], this also coincides with the pain relieving effect which occurs for most patients within 18 minutes.[21] The timeframe was chosen before the data was analyzed.
Secondary outcomes included changes in heart rate, pain scores after epidural activation, neonatal outcomes (Apgar scores at one and five minutes, umbilical venous base excess), and obstetric outcomes (length of labor after epidural, cesarean delivery, mechanical assisted delivery). The statistical assessments for these outcomes were previously reported.[10] Temporal changes in the outcome variables were also assessed for both groups merged, employing the LME model without the adrenaline and time-adrenaline interaction terms. Heart rate values were converted to median values for every five minutes, and thereafter used as the dependent variable in in the abovementioned linear mixed effects model.
The LME analysis was conducted in MATLAB using the fitlme() function from the Statistics and Machine Learning Toolbox. All other statistical calculations were performed in SPSS ® version 24 (IBM, Chicago, IL). A significance level of 5% was used.
At trial conception, there was to our knowledge no data describing continuous hemodynamic changes during labor with the use of epidural in general and no data on the use of epidural adrenaline in particular precluding a reliable power calculation. Based on data from this explorative study a statistical power calculation will define size requirement in a future trial.