Study population

We included patients requiring elective brain tumor surgery at the Department of Neurosurgery, Klinikum Bogenhausen, a 1000-bed teaching hospital of the Technische Universität München, Germany. For perioperative monitoring and maintenance of anesthesia, patients undergoing craniotomy are routinely equipped with a central venous and an arterial line as standard of care in our department. Instead of the regular arterial line, a five french thermodilution catheter (PULSION PVPK2015L20-46N, PULSION Medical Systems AG, Munich, Germany) was placed in a femoral artery at induction of anesthesia and connected to a PiCCOplus thermodilution monitor (Version 7.0; PULSION Medical Systems AG, Munich, Germany).

Patients had to be at least 18 years old and to give informed consent to be included in the study. Exclusion criteria were inability or unwillingness to participate, missing or withdrawn informed consent, chronic atrial fibrillation, and known heart failure or pulmonary disease with dyspnea requiring supplemental oxygen. At study inclusion, the patient's body height and weight were measured.

Thermodilution principle

After injection of a bolus of ice-cold saline through the central venous line into the right atrium, CO is computed from the area under the thermodilution curve obtained by a thermistor at the tip of the arterial catheter [16]. Temporal analysis of the thermodilution curve allows calculation of the central blood volumes [17]. The mean transit time (MTt) is the mean time from the start of injection to detection of the indicator by the arterial sensor, adjusted for recirculation [17]. The downslope time (DSt) describes the exponential decay of the thermodilution curve after its maximum [17]. Multiplication of the MTt with CO equals the total volume marked by the thermal indicator, the intrathoracic thermal volume (ITTV) [17]. Multiplication of the DSt with CO represents the largest compartment of the sequential mixing chambers of the thermal indicator, the pulmonary thermal volume (PTV) [17]. The difference between ITTV and PTV equals the GEDV [1]. ITBV is extrapolated by multiplying GEDV by a fixed factor of 1.25, offering acceptable accuracy in the clinical setting [18]. The difference between the ITTV and the ITBV equals the extravascular lung water (EVLW) [2].

Data acquisition and processing

All monitoring data was stored using the PiCCOWin software (Version 7.0, PULSION Medical Systems AG, Munich, Germany). CO was indexed with body surface area (BSA) calculated from actual body weight and height. GEDV was indexed with BSA using measured height and predicted body weight (PBW), calculated differently for males and females: PBW _male = 50 + 0.91 × (height _cm - 152.4)and PBW _female = 45.5 + 0.91 × (height _cm - 152.4) [19]. BSA was determined by the Du Bois equation: BSA = 0.007184 × length _cm ^0.725 × weight _kg ^0.425 [20]. EVLW was indexed with PBW [21, 22]. These calculations are performed automatically by the PiCCOplus device and not amenable for end user adjustment. From the monitor raw data, MTt and DSt were extracted.

Study protocol

Preoperatively, patients were fasted overnight. Induction and maintenance of anesthesia, surgery and postoperative surveillance on the neurosurgical intensive care unit were performed as per the standards for our department and independent from the study. Thermodilution measurements were performed at least triplicate with 20 ml of iced saline and the mean of a series was taken. The current analysis considers the measurements performed in the morning before discharge of the patients from the ICU on the first postoperative day. Patients not ready for discharge on postoperative day one were excluded.

Study size

Study size was planned using a bootstrapping strategy [23] on our previously analyzed retrospective data [15]. To achieve a power of above 85% for concurrent investigation of the dependencies of GEDV/GEDVI on age and sex, we aimed for analysis of at least 100 patients. This was reached after inclusion of 125 patients between July 2007 and June 2008.

Statistical analysis

For statistical analysis we used the statistical environment R 2.8.1 [24].

The repeatability coefficient was defined as standard deviation divided by the mean of a measurement series.

Univariate analysis was performed using the Kruskal-Wallis-Test for CO, GEDV, GEDVI, MTt and DSt against predefined age groups in decades. The Welch t-test with correction for variance heterogeneity was used to compare the target parameters against gender and to screen for the impact of comorbidities and chronic medication. Gender differences in comorbidities and chronic medication were analyzed using the Chi square test or Fisher's exact test, as appropriate.

As GEDV and GEDVI showed no linear relationship with age, multivariate analysis was applied using generalized additive models [25] with GEDV and GEDVI as targets. Age was fitted with non-linear smoothing, sex as a factor, and height and actual body weight as linear explanatory variables. The combination of single parameters and their possible interactions were compared using significance values and the minimized Akaike Information Criterion [26].

Physiologic rationale

Analysis of the time parameters MTt and DSt from the transpulmonary thermodilution raw data revealed that MTt increases with age, while DSt shows no significant difference. Therefore, the ITTV derived from MTt increases with age and is higher in males. In contrast, the PTV derived from the DSt is independent of age and gender.

The difference between the thermal volumes ITTV and PTV equals the GEDV. Despite its name, the GEDV also includes the volume of the aorta from the aortic valve to the tip of the arterial thermistor [27]. The femoral catheter used in our study, and in most other investigations on transpulmonary thermodilution, has a length of 20 cm. In an adult it is placed with its tip roughly at the iliac bifurcation. It is well known that the aortic diameter increases with age and is larger in males than females [28–32]. Mao and colleagues studied 1442 consecutive asymptomatic subjects scheduled for coronary computed tomography (CT) angiography [28]. Measured with aortic contrast CT, the upper normal limits of the diameter of the ascending aorta were 35.6, 38.3 and 40 mm for females and 37.8, 40.5 and 42.6 mm for males in age groups 20 to 40, 41 to 60 and older than 60 years, respectively. Using an estimated aortic length of 50 cm, a 5 mm increase in luminal diameter would result in approximately 150 ml additional volume. This increase would explain the major part of our findings but does not take into account aortic elongation in elderly subjects [33]. The consequence, again, would be increased distribution volume of the thermal indicator [34].

End-systolic and end-diastolic volumes, measured with cardiac magnetic resonance tomography, are higher in male compared with female patients [35–40]. We also found comparable gender differences in GEDV in our study. Seemingly in contrast to our results, a decrease in cardiac volumes with age is described [35, 37–39]. As we did not perform cardiac imaging in our patients, we are unable to further explain these findings. It is, however, conceivable that the increase in aortic diameter and length may offset the decrease in cardiac volume in older patients.

Indexing problems

Indexing of hemodynamic variables is performed to remove differences between subjects for gender, weight and height. Therefore, theoretically, no significant contribution of any of these factors should persist. However, the contrary is found in our data and the literature [35–41]. Using BSA calculated with PBW for indexing of GEDV, yielding GEDVI, the influence of gender, height and weight remained significant confounders. PBW is dependent on gender and body height [19], but not on actual body weight. The negative correlation of GEDVI with weight in our multivariate analysis suggests that the indexing method overcorrects for heavier subjects.

Likewise, indexing EVLW using gender-specific PBW explains at least part of the higher female EVLWI values in our data. If indexing would be performed equally for both sexes instead of using a gender-specific formula, the difference in EVLWI, but not in GEDVI, would diminish (data not shown). Multivariate analysis with the predictors significant for GEDVI shows that height is negatively correlated with EVLWI, while gender, age and weight have no significant relationships. Therefore, we do not think there is sufficient evidence for a true EVLWI gender difference. The finding in univariate analysis is likely to be related to indexing of EVLW with PBW, which seems overly corrective for larger - more likely to be male - subjects.

Clinical implications

The results of the current study imply that the use of the fixed normal ranges for targeting volumetric therapy is misleading. Although younger patients and females might get severely overhydrated aiming for the proposed normal ranges, older patients may erroneously be deprived of necessary volume. Clinical trials on preload optimization show a lack of consistency on hemodynamic goals and large heterogeneity in treatment effects [42]. Our results may explain part of these findings.

The wide confidence bands for GEDV and GEDVI in our data raise concern of targeting volume resuscitation with absolute values. Relative changes after volume expansion may better indicate volume status [27] and, in our opinion, require further study. If GEDVI changes substantially after a volume challenge, the patient is likely to be volume responsive. In contrast, if the EVLWI shows a pronounced increase, while GEDVI rises only marginally, this may be an indicator of overhydration.

Limitations of the study

The patients in our study appeared to have normal cardiopulmonary function and were evaluated shortly before discharge from the ICU. The overnight fasting required preoperatively is described to have no impact on intravascular blood volume [43]. We cannot exclude a postoperative stress response that may influence cardiac performance or circulating blood volumes. However, a stress response may be present in volunteers or patients examined before induction of anesthesia. Conversely, any premedication may blunt a normal stress level.

Obviously, the patients included in our study were not healthy volunteers. They required craniotomy for removal of a brain tumor. However, we are unaware of any data indicating that patients requiring elective craniotomy present with an abnormal hemodynamic profile. As some of the patients had metastatic brain tumors, we cannot exclude that some of them may have been compromised by their underlying disease. No patient had undergone lobar lung resection, pneumonectomy or chemotherapy at the time of study inclusion.

Our study was not powered to detect a potential impact of chronic medication on preload volumes. However, according to our univariate analysis, the magnitude seems to be far lower than the relationship of preload volumes with age and gender found. In view of the large interindividual variance between subjects, any hypothetical confirmatory trial would have to include at least a ten-fold greater number of patients.

Finally, we did not investigate the influence of interventions such as a volume challenge or passive leg raising on the static preload parameter GEDV. Pulse pressure variation or stroke volume variation are dynamic indicators of cardiac preload and provide valuable information on the volume responsiveness of a patient [44]. Nevertheless, we do think that our findings may help to increase the physiologic understanding of volumetric preload parameters acquired by transpulmonary thermodilution.