Validation of maternal cardiac output assessed by transthoracic echocardiography against pulmonary artery catheterization in severely ill pregnant women: prospective comparative study and systematic review
Department of Obstetrics and Gynecology, Kalafong Provincial Tertiary Hospital University of Pretoria, Pretoria, South Africa
Department of Obstetrics & Gynecology, Erasmus MC University Medical Center, Rotterdam, The Netherlands
Correspondence to: Dr J. Cornette, Department of Obstetrics and Gynecology, Erasmus MC, University Medical Center, Room SP 4460, Dr. Molewaterplein 60, 3015 GJ Rotterdam, The Netherlands (e-mail: firstname.lastname@example.org)
Most severe pregnancy complications are characterized by profound hemodynamic disturbances, thus there is a need for validated hemodynamic monitoring systems for pregnant women. Pulmonary artery catheterization (PAC) using thermodilution is the clinical gold standard for the measurement of cardiac output (CO), however this reference method is rarely performed owing to its invasive nature. Transthoracic echocardiography (TTE) allows non-invasive determination of CO. We aimed to validate TTE against PAC for the determination of CO in severely ill pregnant women.
This study consisted of a meta-analysis combining data from a prospective study and a systematic review. The prospective arm was conducted in Pretoria, South Africa, in 2003. Women with severe pregnancy complications requiring invasive monitoring with PAC according to contemporary guidelines were included. TTE was performed within 15 min of PAC and the investigator was blinded to the PAC measurements. Comparative measurements were extracted from similar studies retrieved from a systematic review of the literature and added to a database. Simultaneous CO measurements by TTE and PAC were compared. Agreement between methods was assessed using Bland–Altman statistics and intraclass correlation coefficients (ICC).
Thirty-four comparative measurements were included in the meta-analysis. Mean CO values obtained by PAC and TTE were 7.39 L/min and 7.18 L/min, respectively. The bias was 0.21 L/min with lower and upper limits of agreement of –1.18 L/min and 1.60 L/min, percentage error was 19.1%, and ICC between the two methods was 0.94.
Las complicaciones del embarazo más graves se caracterizan por trastornos hemodinámicos serios, debido a los cuales existe la necesidad de sistemas validados de monitorización hemodinámica para mujeres embarazadas. Aunque la cateterización de la arteria pulmonar (CAP) mediante termodilución es el patrón de referencia clínico para la medición del gasto cardíaco (GC), este método se usa con poca frecuencia debido a su naturaleza invasiva. La ecocardiografía transtorácica (ETT) permite la determinación no invasiva del GC. El objetivo de este estudio fue validar la ETT frente al CAP para determinar el GC en mujeres embarazadas gravemente enfermas.
Este estudio consistió en un metaanálisis que combinó datos de un estudio prospectivo y una revisión sistemática. El estudio prospectivo se llevó a cabo en Pretoria (Sudáfrica) en 2003. Se incluyeron mujeres con complicaciones graves en el embarazo que requerían una monitorización invasiva mediante CAP según las directrices de ese momento. Se realizó una ETT en un plazo de 15 minutos de haber realizado el CAP y el investigador no tuvo acceso a las mediciones del CAP. Las mediciones comparativas se extrajeron de estudios similares obtenidos a partir de una revisión sistemática de la literatura y se añadieron a una base de datos. Se compararon las mediciones simultáneas del GC mediante ETT y CAP. La concordancia entre métodos se evaluó a través del método estadístico de Bland-Altman y de coeficientes de correlación intraclase (CCI).
Se incluyeron treinta y cuatro mediciones comparativas en el metaanálisis. Los valores medios del GC obtenidos mediante CAP y ETT fueron de 7,39 l/min y 7.18 l/min, respectivamente. El sesgo fue de 0,21 l/min, siendo los límites inferior y superior de la concordancia de −1,18 l/min y 1.60 l/min; el error porcentual fue del 19,1%, y el CCI entre ambos métodos fue de 0,94.
Las mediciones del GC en mujeres embarazadas mediante ETT muestran una excelente concordancia con las obtenidas mediante CAP. Dada su naturaleza no invasiva y su disponibilidad, la ETT podría considerarse como referencia para la validación de otras técnicas relacionadas con el GC en mujeres embarazadas.
Pregnancy imposes a substantial challenge on the maternal cardiovascular system[1, 2]. Complications such as pre-eclampsia, cardiac disease, sepsis, hemorrhage and pulmonary embolism, which account for the majority of severe maternal morbidity and mortality, are characterized by profound hemodynamic disturbances[3-6]. Maternal pulse rate and blood pressure are easily obtained and are often used alone as indirect surrogates of maternal cardiovascular function. Nevertheless, knowledge about cardiac output (CO) can be important when managing hemodynamically compromised pregnant women or studying (patho)physiological conditions in pregnancy[3, 4, 7-12]. Thermodilution by pulmonary artery catheterization (PAC), using a Swan–Ganz catheter, is considered the clinical gold standard for CO measurement. Until the early 2000s, it was commonly used for hemodynamic monitoring and guiding therapy in intensive care settings[14, 15]. In pregnancy, critically ill and severely pre-eclamptic women were also managed with PAC[16-25]. Nevertheless, this invasive technique requires right-heart catheterization with its inherent procedure-related risks[26-29]. Controversy with the technique began after several reports failed to show any benefit or even suggested increased mortality with its use in various critical conditions[13, 30-36]. The initial enthusiasm for this technique faded in the intensive care, and subsequently in the obstetric, community, leaving a gap in hemodynamic monitoring which has not yet been filled by new minimally- or non-invasive alternatives, as validation of these methods remains a concern.
Transthoracic echocardiography (TTE), using two-dimensional (2D) and pulsed-wave Doppler of the left ventricular outflow tract (LVOT), is commonly used to determine CO in both pregnant and non-pregnant women. The technique is non-invasive, safe and accessible to pregnant women, as many obstetric ultrasound devices allow upgrading with cardiac software and probes. Nevertheless, validation in pregnancy against the clinical gold standard of PAC has not been performed adequately. As indications for PAC in pregnancy are limited to severely ill women, comparative studies included limited numbers of subjects and were often performed using statistical methods that are nowadays considered suboptimal or inappropriate. By combining data from a single-center comparative study with those from a systematic review of the literature in a meta-analysis, our aim was to validate the determination of CO using TTE against PAC in pregnant women.
The prospective comparative trial was conducted at the Kalafong Hospital, a tertiary care referral center for the University of Pretoria in South Africa, from May–October 2003. Severely pre-eclamptic women, admitted to the obstetric high-care unit and requiring PAC for their clinical management, were included in the study after informed consent had been obtained. The study was approved by the medical ethics board of the University of Pretoria (40-2003). Indications for PAC were according to contemporary guidelines[16, 18, 21, 25]. These recommended PAC in severe pre-eclampsia complicated by oliguria (not responding to fluid challenge), severe hypertension (not controlled by a combination of three different antihypertensive drugs), pulmonary edema or by clinical or echocardiographic signs of cardiac dysfunction. A triple lumen continuous cardiac output PAC (7.5F) (Edwards Life Sciences, Irvine, CA, USA) was inserted via the internal jugular vein approach. Correct position was confirmed by waveform analysis and chest X-ray. This type of catheter allows both intermittent and continuous CO determination using a vigilant CO computer (Edwards Life Sciences).
Intermittent CO determination obtained by bolus thermodilution is considered the clinical gold standard, and was used for comparison with TTE. Measurements were performed soon after inclusion in the study and insertion of the PAC. CO was calculated from the mean of three consecutive thermodilution curves using 10 mL physiological saline room temperature injectates at different phases of the respiratory cycle. Subsequent CO measurements were performed with the continuous CO module; a 10-cm thermal filament is incorporated into the pulmonary artery catheter, 15–25 cm proximal to the catheter tip, which emits pulses of energy and thereby heats blood in a repetitive intermittent sequence. Differences in temperature measured by the thermistor at the catheter tip are correlated with the emitted signal. CO is determined by a similar equation as for thermal dilution without the need for repetitive fluid injections. A continuous output, which is an average of the CO measured over the previous 5–15 min, is deduced[13, 37]. Clinical management was based on continuous CO measurements and cardiac and pulmonary pressure readings.
Subsequently TTE was performed by the principal investigator using an obstetric ultrasound system with appropriate cardiac transducer and software package (Siemens Sonoline Omnia, Siemens, Munich, Germany). The LVOT diameter (LVOTd) was measured at the base of the aortic leaflets from a parasternal long window, from which the LVOT cross-sectional area (LVOTcsa) was calculated (0.7854 × (LVOTd)). The LVOT velocity–time integral (LVOTvti) was obtained by pulsed-wave Doppler from an apical five-chamber view, and stroke volume (SV) was computed as LVOTcsa × LVOTvti. CO was calculated by multiplying SV by the corresponding heart rate derived from the simultaneous electrocardiography recordings. The mean of three measurements was used for analysis. Measurements were recorded and calculations of LVOTcsa, SV and CO were performed offline after completion of the TTE examination.
Both pulmonary artery thermodilution and echocardiography Doppler measurements were performed in a 15° left lateral tilt to limit the possible interference of aortocaval compression. PAC and TTE measurements were performed within 15 min of one another without the occurrence of new therapeutic interventions or major clinical changes between the two measurements. The investigator performing the ultrasound measurements was blinded to the thermodilution measurements.
A systematic review of the literature was performed. We searched EMBASE, MEDLINE, Web-of-Science, Scopus, the Cochrane Library, CINAHL, PubMed and Google Scholar with search headings such as pulmonary artery catheter, echocardiography, Doppler and pregnancy. The search strategy for each database is provided in Appendix S1. Reference lists of relevant articles were screened for potential additional records not uncovered by the search strategy.
Articles in English describing direct comparison of CO determined by TTE using 2D and pulsed-wave Doppler at the LVOT and PAC using bolus thermodilution, during or immediately after pregnancy, were included. Studies using a different reference method (e.g. Fick method) or different ultrasound method (e.g. continuous-wave Doppler in the aortic, mitral or pulmonary position) were excluded. The methodology employed was in accordance with the PRISMA statement. Titles and abstracts identified from the search were screened and full-text reports were obtained and analyzed for studies that seemed to be relevant. Individual comparisons between both methods were extracted from articles that met the predefined criteria.
A database was created containing direct comparisons between the TTE and PAC data obtained during the prospective arm of the study combined with data obtained from the systemic review. Agreement between both methods of CO measurement was evaluated with Bland–Altman plots and statistical analysis as appropriate for both the prospective arm of the study and the combined database. Mean CO, bias, SD around the bias, limits of agreement (mean CO ± 1.96 SD around the bias) and percentage error ((1.96 SD around the bias/mean CO) × 100) were determined. Agreement was considered to be good if bias was low and percentage error was below 30%, as proposed by Critchley and Critchley. Absolute agreement in ratings was evaluated using the intraclass correlation coefficient (ICC).
Seven severely pre-eclamptic women were included in the prospective arm of the study, an abstract for which has been published previously. The demographic characteristics, indications for right-heart catheterization and time of measurement (antepartum or postpartum) are presented in Table 1. No maternal mortality occurred. One neonate was born at 25 weeks' gestation and died immediately postpartum after comfort care was offered, as birth at this gestational age was considered non-viable in South Africa. No clinically relevant complications related to the PAC occurred, with the exception of a balloon rupture in one woman. The problem was quickly suspected after insufflation of the balloon with air failed to produce a typical wedged waveform. Replacement of the catheter over a guidewire confirmed this suspicion. Close observation did not reveal any clinical signs of air embolism. Mean ± SD CO obtained by thermodilution and Doppler echocardiography were 6.89 ± 2.17 L/min and 6.46 ± 1.84 L/min, respectively. Mean ± SD CO was 6.67 ± 1.99 L/min, with a bias of 0.43 L/min and SD around the bias of 0.63 L/min. The limits of agreement were –0.8 to 1.66 L/min. Percentage error was 18.4%. ICC was 0.97 (95% CI, 0.79–0.99).
Table 1. Demographic characteristics and indication for right-heart catheterization in 34 pregnant women with cardiac output (CO) measurements reported among three studies
The electronic search identified a total of 366 records; one additional record was identified through reviewing the reference lists of those identified from the electronic search (Figure 1). After duplicate records had been removed, 220 records were screened on the basis of title and abstract, and 25 full-text articles were assessed for eligibility. Of these, 10 included a direct comparison between PAC and echocardiographic measurements in general during pregnancy[40-49]. Seven described comparative measurements of CO specifically, of which four were excluded because the methods of CO determination were different from our predefined methods. Finally, three articles, including an abstract of the current prospective study, were retained for direct comparison and meta-analysis[40, 41, 44]. One study, by Lee et al., included 16 subjects. One woman had been excluded from the study owing to a technical difficulty resulting from a faulty PAC. The article included a table with comparative measurements in each subject from which the data were extracted. The other study, by Belfort et al., included 11 women. No complications related to right-heart catheterization were reported. Datapoints were derived from the regression lines presented in the article.
Measurements obtained by TTE and PAC from our study were added to the data retrieved from the systematic review. All three studies in the review reported a single paired measurement per subject. In total, 34 subjects were included in the meta-analysis, of whom 24 were examined antepartum and 10 postpartum (Table 1). The Bland–Altman plot presented in Figure 2 suggests good agreement between the methods over a wide range of CO measurements (3.95 to 11.40 L/min). Mean CO obtained by thermodilution and Doppler echocardiography were 7.39 ± 2.09 L/min and 7.18 ± 2.10 L/min, respectively. Mean CO was 7.28 ± 2.07 L/min, with a bias of 0.21 L/min, and SD around the bias of 0.71 L/min. Limits of agreement were –1.18 to 1.60 L/min. Percentage error was 19.1% and ICC was 0.94 (95% CI, 0.88–0.97).
Maternal CO is an important hemodynamic parameter that is subjected to substantial changes in pregnancy[1, 50]. TTE using pulsed-wave Doppler ultrasound in the LVOT position is commonly used for CO measurement[2, 3, 8, 12, 51]. Nevertheless, our systematic review revealed only three validation studies, including our own prospective study, of limited size, comparing TTE with the clinical gold standard[40, 41, 44]. By combining data from these studies we were able to analyze 34 paired measurements using appropriate statistics. These data covered a wide range of CO and gestational and maternal ages, both during pregnancy and immediately postpartum, in different pathological conditions and in three independent research groups. Our results show excellent agreement with small bias, limits of agreement and percentage error well within our predefined margins and an excellent ICC between the two methods. When a new method for assessing a clinical variable is introduced it is usually compared with an established reference technique, and adoption of the new method usually depends on the degree of agreement with the reference technique and other potential benefits.
PAC is an invasive technique with inherent risks that have been well described both in and outside pregnancy[26-29, 53, 54]. In our systematic review, we encountered two complications related to the procedure (balloon rupture and faulty PAC). TTE is non-invasive and increasingly accessible, as many obstetric ultrasound devices allow upgrading with cardiac software and probes. It can be used in all pregnant women, from healthy to critically ill, at the bedside. As fetal and adult echocardiography are, in essence, very similar, it would take most fetomaternal medicine specialists little effort to learn the appropriate planes. By analogy with its homonyms in other critical circumstances (FAST, BLEEP, FATE or HART) a ROSE (rapid obstetric screening echocardiography) scan can be developed with obstetric anesthesiologists and congenital cardiologists for rapid, accessible and now reliable bedside hemodynamic monitoring in pregnant women.
Agreement between two different techniques depends on the accuracy and precision of both methods. Accuracy describes how close the measurement is to the reference value and precision describes how close the values of repeated measurements are to each other. Cecconi et al. nicely interpreted this definition by comparing it with target shooting, in which accuracy is the characteristic of being able to shoot close to the bullseye and precision is how close repeated shots are to each other. Most studies performed in non-pregnant women and the two studies included in our meta-analysis compare the two techniques by correlation and regression. However, these merely reflect the strength of a relationship and not the agreement. If, for example, each CO value determined by TTE was exactly 5 L/min higher than the one determined by PAC, the standard Pearson's coefficient would still show perfect correlation despite substantial differences between the two techniques. By centering and scaling data using a pooled mean and SD, the ICC is more appropriate for reflecting variance and agreement between two methods of measurement.
In their reference papers, Bland and Altman[56, 57] proposed bias and precision statistics to analyze agreement between two methods for which the differences (bias) are plotted against the means of each pair of measurements. This is now considered the gold standard for comparing two techniques of CO measurement. The bias, reflecting accuracy, and the SD around the bias, as well as limits of agreement (limits in which 95% of datapoints fall on each side of the bias), estimating precision, can be calculated. In order to conclude that agreement between two methods is acceptable, it should be decided beforehand where the limits of agreement should fall. It is often difficult to determine which limits are clinically acceptable. The reference method of thermodilution is merely the clinical standard and does not reflect the true CO. It carries some inherent errors in accuracy and precision (10–20%) due to fluctuations in CO with respiration and technical limitations[13, 37]. It is also evident that a bias of 1 L/min is more significant at a low CO (e.g. 3 L/min) than in a high output state (e.g.10 L/min). To overcome these problems, Critchley et al. proposed that a new method should be acceptable if the level of accuracy and precision is at least equal to that of the reference method. They proposed that the percentage error of the limits of agreement as compared with the mean be used to assess agreement between two methods of CO determination, with a cut-off of 30%[39, 52, 58]. Studies comparing both methods of CO determination in non-pregnant adults using the appropriate Bland–Altman statistics are equally rare; they are of similar size and suggest a similar degree of agreement[59-61].
The importance of CO measurement in obstetrics is highlighted by the emergence of a multitude of new techniques for assessing CO in a minimally- or non-invasive way. Several devices calculate CO using pulse contour analysis (LidCO®, Nexfin®), impedance cardiography (Niccomo®, Physioflow®, NICOM®) or continuous-wave ultrasound (Uscom®)[7, 62-65]. Some of these allow continuous measurement and some require only a limited amount of training and skill, making them more operator independent. Nevertheless, validation of these devices, especially in pregnancy, remains a major issue, mainly owing to the lack of a gold standard.
PAC is still considered the clinical gold standard for CO measurement, but indications for its use in obstetrics have become extremely rare. Our study showed that TTE is equivalent to PAC for CO measurement in pregnant women and, aside from its clinical use, can be considered a surrogate gold standard for validating other techniques.
The strength of this study lies in the fact that by the systematic review approach we were able to gather a maximal number of cases for analysis using the appropriate statistics. The main limitations are that 34 datasets is quite a modest number and that our study did not include healthy pregnant women.
In conclusion, our data indicate that CO measurements by TTE agree with those obtained by PAC in severely ill pregnant women. Given its non-invasive nature and availability, TTE could be considered as a reference for the validation of other hemodynamic measurement techniques in pregnant women.
We thank Wichor Bramer of the Erasmus MC Medical library for his expert help with the systematic review of the literature. We thank Edwards Lifesciences, South Africa, who donated the PACs, Vigilant computers and pressure transducers for this study as well as Siemens, South Africa, who kindly put the cardiac ultrasound transducer and software at our disposal for the study. The research project was funded by a grant from the Maternal and Infant Health Care Strategies Research Unit (MRC) of South Africa.