Assisted Fluid Management for Major Liver Surgery: Reply

In Reply:

We received interesting comments^1–5 regarding our recent article on the use of a decision support fluid therapy system during major liver resection surgery.⁶ After considering the queries, we reviewed our patient records to provide clear answers to these important comments.

Regarding the preoperative hepatic function of our patients, none had liver insufficiency, as this is a contraindication to major hepatic resection. Preoperative indicators of liver injury, such as mean ± SD aspartate aminotransferase and alanine aminotransferase did not significantly differ between groups (38 ± 16 vs. 35 ± 13 units/l for the decision support and control group [P = 0.294] and 38 ± 23 vs. 36 ± 25 units/l [P = 0.703]). We also investigated indicators of liver function, including the international normalized ratio, bilirubin, factor V, and albumin. All were nonsignificantly different between groups. Moreover, there was no significant difference between groups regarding the number of patients under metformin preoperatively (two patients in the control group vs. three in the decision support group; P > 0.999).

Inquiring about the type of administered crystalloid is also interesting. The only crystalloid we used was Ringer’s lactate. Although it does contain lactate, this had negligeable impact on lactate concentrations in our study. This is clearly demonstrated by the fact that less Ringer’s lactate was administered in the restrictive group up to the end of hepatic resection, and despite this, this group had higher lactate values across the different time points. The only time point when lactate was similar between groups was at baseline, right after induction of anesthesia.

Two patients died in our study, and both were in the restrictive group, leading to a 4% mortality in this group. The first patient died 3 weeks after surgery on the floor from a pneumonia due to aspiration of gastric contents leading to a cardiac arrest and the death of the patient. The second patient passed away 15 days after surgery. She had an urgent redo surgery on postoperative day 1 for a portal thrombosis, which led to a hepatic ischemia, hepatocellular insufficiency, and later to multiorgan failure. These two events were unlikely linked to the fluid strategy applied intraoperatively. All the surgeons in our center (Paul Brousse Hepatobiliary Center, Villejuif, France) are specialists in liver surgery. The Paul Brousse Hepatobiliary Center is a highly specialized center dedicated to liver and pancreatic surgeries. It is the biggest center in France for liver transplantation and the third largest in Europe with more than 160 liver transplants per year and many more liver resections. As such, a negative impact of a surgeon’s experience on patient outcome is highly unlikely. Additionally, even if pulse contour technology may have some well-known limitations, none of our patients had preoperative right heart dysfunction (cardiac echo done in all patients preoperatively and all normal).

Another important comment to which several readers alluded is the correlation between lactate and norepinephrine dose. We fully agree with this and believe that one of the reasons why lactate was lower in the decision support group is because the assisted fluid management (Edwards Lifesciences, USA) software that guided fluid therapy in that group was able to optimize preload without excess.⁷ This consequently increased cardiac output without a clinically significant increase in central venous pressure. The ultimate effect was probably increased tissue perfusion, as has been shown in a recent randomized trial that studied the impact of the assisted fluid management software on microcirculation during abdominal surgery.⁸

Finally, we appreciate the readers’ comment regarding potential residual imbalances despite randomization. While randomization is expected to balance confounders on expectation, minor imbalances can still occur by chance. There is ongoing debate regarding the appropriate standardized mean difference threshold for defining meaningful imbalance (e.g., 0.1 vs. 0.2). Given this, we conducted an adjusted analysis to assess whether variables with observed standardized mean differences greater than 0.1 influenced our findings.

In the unadjusted analysis, the median (quartile 1, quartile 3) lactate level in the decision support group was 2.5 (1.9, 3.7) and 4.6 (3.1, 5.4) in the restrictive groups. A Mann–Whitney U test confirmed a statistically significant difference (P < 0.001), as reported in the article. The percentage reduction in median lactate was consequently 45.7% (calculated as [2.5 – 4.6] divided by 4.6, multiplied by 100). Because our adjusted analysis required a regression approach, we first applied a log transformation to lactate to better meet model assumptions. In the unadjusted regression model, the coefficient estimate for the decision support group was −0.512 on the log scale, corresponding to a 40% lower lactate level during decision support fluid therapy when compared to a restrictive fluid strategy (because e to the power of −0.512 is approximately 0.60). For the adjusted analysis (including all table 1 variables with standardized mean difference greater than 0.10), the coefficient for the decision support group was −0.444, translating to a 36% lower lactate level compared to the restrictive fluid therapy group (because e to the power of −0.444 is approximately 0.64). Importantly, the group effect remained highly significant (P < 0.001) in both adjusted and unadjusted models, demonstrating that potential residual imbalances did not influence the primary result.

To recapitulate on the query regarding standardized mean difference, we demonstrate that across all three methods—difference in medians, unadjusted regression (log-transformed), and adjusted regression (log-transformed with all variables with a standardized mean difference greater than 0.10)—the percentage reductions were of similar magnitude (45.7, 40, and 36%, respectively), and all were highly statistically significant (P < 0.001).

In conclusion, the readers asked interesting questions regarding the baseline characteristics of the patients included in our study. We investigated these queries and found no significant difference in indicators of acute liver injury (aspartate aminotransferase and alanine aminotransferase), indicators of liver dysfunction (international normalized ratio, bilirubin, albumin, and factor V), or metformin use. Furthermore, we provided evidence that the use of Ringer’s lactate was not the cause of the observed hyperlactatemia. Additionally, we further investigated the causes of mortality and considered the readers’ comments regarding surgeon experience and limitations of pulse contour analysis. Finally, we carefully investigated our statistical methodology and provided different analytical perspectives to demonstrate that indeed, the patients in both groups had no clinically relevant preoperative differences other than the study intervention.