Leakage of albumin in major abdominal surgery

Patients (n = 12) scheduled for major pancreatic (n = 9) or oesophageal surgery (n = 3) were recruited. One patient was excluded from further analysis because of an inoperable tumor and therefore a short surgical procedure, and one patient because exogenous albumin was given as colloid during surgery due to elevated plasma creatinine (presented as a case report, see Additional file 1: Text S1 and Additional file 2: Figure S1). The remaining 10 patients (2 females) were aged 65.7 ± 5.6 years and had a body mass index of 24.0 ± 2.3 kg/m². The patients had a number of co-morbidities apart from their tumors: hypertension 4, diabetes mellitus 3, previous myocardial infarction 1, previous chemotherapy 2, congestive heart failure 2, atrial fibrillation 2, hypothyreosis 2, and one each of pancreatitis, trachea-esophageal fistula, multiple myeloma, multiple sclerosis, and deep venous thrombosis. The attending anesthesiologist graded preoperative health according to the American Society of Anesthesiologists’ (ASA) classification of physical health, all within grade II (n = 6) or III (n = 4). Two patients had a preoperative body weight loss of –6.8 and –10.0 kg over 1 and 2 months, respectively, corresponding to a weight loss of more than 10 % and a nutritional risk score of five [7]. The body weight of the remaining eight subjects differed by –0.6 kg (–1.8 to 0.5) from values 1–2 months preoperatively.

P-alb, blood hematocrit (B-Htc), and blood hemoglobin (B-Hb) were measured 15 times over 3 days according to a predefined protocol. The first sample was taken as soon as an arterial line was established, then after induction of anesthesia, at the start of surgery, after 1 h of surgery, at the end of resection, at the start of reconstruction, at the end of surgery, 1, 2 and 3 h after surgery, on postoperative day 1 at 8 a.m., 10 a.m. and 4 p.m., on postoperative day 2 at 8 a.m., and on postoperative day 3 at 8 a.m.. Losses of albumin and hemoglobin due to bleeding, by suction and sponges, and postoperatively by drains were assessed by weighing and measurement of hemoglobin and albumin content. Losses of albumin in urine were considered as insignificant. Data from all intravenous infusions were collected. B-Hb was also measured in transfused donor blood and P-alb in donor plasma.

Anesthesia was performed according to the unit routines at the time of the study. Briefly, all patients had an arterial line, central venous catheter, thoracic epidural block with bupivacaine, fentanyl and epinephrine, and general anesthesia that after induction was maintained by sevoflurane in oxygenated air at an age-corrected minimal alveolar concentration of 0.8. Ventilation was adjusted to achieve normocapnia based on blood gas analyses. Intravenous fluids comprised starch (2 mL/kg/h; Volulyte®; Fresenius Kabi, Uppsala, Sweden), acetated Ringer’s solution (2 mL/kg/h), and glucose (25 mg/mL, 1 mL/kg/h) according to the unit’s routine. In this pragmatic study, bleeding was treated at the discretion of the attending anesthesiologist.

P-alb was analyzed by the Study Center at the Karolinska University Laboratory using nephelometry with a coefficient of variation of 1.9 % (IMMAGE® 800; Beckman Coulter AB, Bromma, Sweden), whereas B-Hb and B-Hct were assessed on a blood gas analyzer (ABL800 FLEX; Radiometer Medical ApS, Brønshøj, Denmark) at the Department of Anaesthesia.

Baseline blood volume (BV) was calculated anthropometrically from gender and body size [8], and plasma volume (PV) was derived by:

The f-ratio 0.91 represents the ratio between total body hematocrit and large vessel B-Hct. The intravascular hemoglobin mass (MHb) is then:

By combining (1) and (2) we achieve:

Intravascular albumin mass (IAM) can be similarly calculated by:

Equations 1–4 are valid at all time points. When considering consecutive time points (n + 1 versus n) in the predefined protocol, MHb at time n + 1 can be calculated from the value at time n and measured loss and gain of hemoglobin in that time interval according to:

When this MHb _n+1 from equation (5) is inserted into equation (3) together with B-Hb _n+1 and B-Hct _n+1 it is possible to calculate PV _n+1. This value can be inserted into equation (4) together with P-alb _n+1 generating IAM _n+1, representing albumin mass at time point n + 1 related only to B-Hb and B-Hct and measured loss and gain of hemoglobin. IAM’ represents another way to assess mass balance of albumin—directly by considering losses and gains of albumin over time. The apostrophe denotes that these IAM’ values are obtained differently. Gains are estimated from albumin content in plasma transfusions and albumin infusions, losses from measurement of albumin in drains or from estimated bleeding in suction bottles and sponges according to:

The cumulative difference between these two measures over time is presented in this paper as the cumulative perioperative albumin shift, supposedly to the extracellular space:

Finally, the fractional plasma volume dilution at time n (fPVdil _n ) is then related to the baseline plasma volume PV₀ according to:

Similar fPVdil calculations serve as in-data in volume kinetic modelling [9].

In a previous study of similar patients investigated under the same unit routines, the drop in P-alb from start of surgery to the second postoperative day was –11.5 ± 2.6 g/L [6]. This large and consistent drop, corresponding to an effect size of 4.4 (the difference divided by the standard deviation of the difference), made it likely that sufficient information of the time course would be obtained with a sample size of 10.

P-alb decreased (p < 0.0001; Fig. 1a), mainly during the first part of the surgical procedure, from 32.8 ± 4.8 g/L at baseline to 18.7 ± 4.8 g/L (p < 0.001) at the start of reconstructive surgery, and to 19.8 ± 3.4 g/L (p < 0.001 compared with baseline) 3 days postoperatively (Fig. 1a) in subjects not treated with a continuous albumin infusion (n = 10). At the end of surgery, P-alb had decreased by 15.1 ± 5.2 g/L (95 % confidence interval of the mean decrease 11.4–18.8; p < 0.001) corresponding to 45 ± 12 % of the baseline value. Notably, there was a significant decrease in P-alb already at the start of surgery, as compared to baseline before the start of anesthesia.

Plasma volume increased (p = 0.0006; Fig. 1b) during surgery to reach its maximum at the start of the surgical reconstruction, but was postoperatively not different from baseline (Fig. 1b). Cumulative perioperative albumin shift changed (p < 0.0001; Fig. 1c) and the mean value increased throughout surgery to reach the most pronounced difference from baseline 1 h after the end of surgery (24 ± 17 g; 95 % confidence interval of the mean 11–36; p < 0.001), but from postoperative day 1 onwards did not differ from baseline (Fig. 1c). In total, six of the 10 patients were given 11 units of plasma during the study period containing a total of 88.2 g albumin. Four of those were given during the surgical procedure. Another 125 g albumin was given as albumin infusions, leaving only two subjects without any exogenous administration of albumin. Measured gains and losses of albumin are presented in Table 1. Similarly, in total, 15 units of erythrocyte concentrate with a mean hemoglobin of 192 g/L and a mean hematocrit of 0.59 were given to six different patients during the study period. The mass balance of albumin for four time periods is presented in Table 1, and cumulative albumin balance in Table 2. Laboratory tests are presented in Table 3.

Surgery time was 369 ± 84 min. Further details of fluid balance during and after the surgical procedures are given in Tables 1 and 2. Five patients had pancreatoduodenectomy, one of these also had splenectomy and vascular reconstruction; one patient had distal pancreatic resection with hemicolectomy and splenectomy, one patient had pancreatic resection with resection of the upper mesenteric artery, one patient had reconstruction of the oesophagus, one patient had transthoracic resection of the oesophagus, and one patient had transhiatal resection of the oesophagus. Three patients (oesophageal resection in two and pancreatic resection in one) started with enteral feeding on postoperative day 1, whereas the other patients remained fasting, only drinking small amounts of water.

Gains and losses of fluid and the resulting fluid balance over four time periods during 3 perioperative days are also presented in Table 1. Cumulative fluid balance changed over time (p < 0.0001; Table 2) and, when consecutive time points were analyzed simultaneously, both preoperative versus end of surgery, and end of surgery compared with postoperative day 1 differed significantly (p < 0.01), even after Bonferroni correction as presented in Table 2. On postoperative day 2 body weight gain reached its highest value of 3.8 ± 1.7 kg (p < 0.001; Table 2), and cumulative fluid balance was then 5.4 ± 2.0 L (p < 0.001; Table 2). At the same time point the correlation between fluid balance and weight gain did not reach statistical significance (p = 0.091, r ² = 0.316), neither did cumulative perioperative albumin shift correlate with body weight gain (p = 0.336, r ² = 0.12) or fluid balance (p = 0.206, r ² = 0.19).

The mechanisms of the P-alb decrease during surgery are unclear; they are most likely multifactorial, and might even be variable over time. The rapid fall in P-alb before surgery might be related to general sympathetic tone, posture, vasodilation caused by the epidural block, or effects on blood pressure attributed to general anesthesia. There is an extravasation of albumin and other plasma proteins in anesthetized or septic animals of different species after infusion of a variety of different fluids such as whole blood [12], albumin [13], and crystalloids [14]. Three major reasons behind this extravasation are discussed: overhydration, increased capillary leakage, and change in albumin turnover.

The changes in intravascular albumin mass were calculated from changes in hemoglobin concentration, and demonstrated to be in accordance with plasma protein loss after major gynecological surgery [23]. However, the method used has been criticized for insufficient precision in determination of albumin balance which will propagate to the calculation of cumulative perioperative albumin shift. To determine the possible impact of this level of precision, simulations were performed over possible confounders. Anthropometric blood volume calculations, uncertainty in the f-ratio between total body hematocrit and B-Hct, variation of this ratio over time, the difficulty in accurately assessing the losses of hemoglobin in the clinical setting of major abdominal surgery, and the impact of uncertainty in P-alb and B-Hb measures were evaluated. These five possible sources of error influence the magnitude of the changes of the albumin shift versus time plot, but not the direction or time pattern (see Additional file 3: Text S2; Additional file 4: Figure S2; Additional file 5: Figure S3; Additional file 6: Figure S4).

Apart from small analytical errors of P-alb, short-term effects on P-alb (minutes) are likely associated with simultaneous changes in plasma volume evoked by changes in body position [24], or states of under- or overhydration. Exogenous vasopressors such as norepinephrine are associated with plasma volume contraction [25], and we speculate that physiological stress can also have similar rapid effects. Intermediate effects on P-alb (hours) include compromised lymphatic return of albumin, either by para-aortal resection of lymph glands in major abdominal surgery or by altered lymphatic outflow pressure (i.e., central venous pressure) caused by positive pressure ventilation [26], the state of hydration, or posture [27]. Also, the transcapillary escape rate can be increased by overhydration or inflammation [4]. Long-term changes in P-alb (days to months) can be associated with increased synthesis (steroids), decreased synthesis (malnutrition or liver failure) [28], increased losses (inflammatory bowel disease or renal failure) [28], or increased volume of distribution (obesity or ascites). However, these long-term factors are less likely to affect the drop in P-alb over a couple of hours described in this paper.

The strength of this pilot study lies in the presentation of new knowledge on the time course of albumin leakage during major abdominal surgery while keeping track of gains and losses. The robustness of the results in a small group of subjects is demonstrated by simulation over confounders. Among limitations of the study are the limited external validity related to the utilized fluid routine with starch (now abandoned), that blood substitution was left to the discretion of the attending anesthesiologist, and the difficulty in assessing bleeding correctly in such a complicated situation as open abdominal surgery. Still, the temporal aspects of capillary leakage, as reflected by the albumin concentration and the albumin shift, are important and need to be considered when perioperative fluid regimens are studied. Here, our study adds original information.