Introduction

Pylorus-preserving pancreaticoduodenectomy (PPPD) is one of the most invasive intraabdominal surgeries; it requires a prolonged recovery time after surgery because of its extensive incision, invasiveness, severe postoperative pain, decreased lung function, and postoperative complications [1]. Therefore, it is important to seek out every available opportunity to speed up recovery in order for patients to return to their daily lives as soon as possible after the operation.

Hypocapnia caused by hyperventilation during general anesthesia is a common situation. Traditionally, it has been traditionally advocated, to some extent, because it has been considered to prevent the restoration of spontaneous respiratory efforts and, thus, may reduce the need for muscle relaxants and additional anesthetics [2]. However, hypocapnia has several possible disadvantages in terms of maintaining a normal physiology as it can reduce cerebral blood flow and oxygen delivery [3]. Hypocapnia also causes an increase in the capillary permeability of the lungs, bronchoconstriction, an increase in shunt as a result of decreased lung compliance, and a leftward shift of the hemoglobin dissociative curve [4]. Moreover, it reduces splanchnic blood flow and increases myocardial oxygen demand, hypercoagulability, and the incidence of dysrhythmias [4]. Although the arterial partial pressure of carbon dioxide (PaCO2) is not monitored continuously, its status is well reflected by end-tidal carbon dioxide (EtCO2) levels, which are routinely monitored during anesthesia [5]. Despite these shortcomings, EtCO2 tends to remain as low as ever during surgery [2], and poor clinical outcomes related to hypocapnia have not been sufficiently studied [6, 7].

Prior studies have shown that intraoperative hypocapnia is related to worse outcomes in patients undergoing lower abdominal surgery [6, 7]. However, it is unknown whether deleterious effects from hypocapnia exist in other surgeries, such as PPPD. We hypothesized that hypocapnia in patients undergoing anesthesia for PPPD is associated with worse clinical outcomes compared with patients maintained under normocapnia or hypercapnia.

In this retrospective study, we investigated the effect of intraoperative hypocapnia on hospital stay and mortality in patients undergoing PPPD.

Materials and methods

Ethics

This study was approved by the institutional review board (IRB) of Severance Hospital (no. 4-2019-1164), and the need for informed consent was waived.

Subjects

Data were retrospectively collected from patients who underwent PPPD for a suspicious malignancy at a tertiary university hospital (Severance Hospital) from January 2006 to December 2015. Patients were excluded if they expired during surgery, underwent surgery assisted by laparoscopy or robot, or had undergone cooperation or reoperation.

The patients were divided into two groups based on their mean EtCO2 values during anesthesia, namely the hypocapnia group (less than 35 mmHg) and the normocapnia group (equal to or greater than 35 mmHg), as proposed by Way and Hill [8].

Perioperative variables

We collected data related to subject age, sex, height, weight, body mass index (BMI), comorbidities, American Society of Anesthesiologists (ASA) classification, tumor location, and malignancy. The collected intraoperative variables were the duration of operation, duration of anesthesia, and peak inspiratory airway pressure (PIP), and EtCO2, which was recorded every 15 min in the anesthesia records. Hemodynamic data including heart rate, systolic blood pressure (SBP), diastolic blood pressure (DBP), and central venous pressure (CVP) were recorded every 5 min during surgery. Laboratory data, including PaCO2 from arterial blood gas analysis (ABGA), were also recorded every hour during the operation. The mean values of the variables collected during operation were used for analysis.

Outcomes

The primary outcome was the comparison of postoperative length of hospital stay (LOS) after surgery between the groups. Secondary outcomes included the length of ICU stay; postoperative 30-day, 1-year and 2-year mortality; and incidence of postoperative pancreatic fistula (POPF) as the major postoperative complication [9].

Statistical analysis

Statistical analyses were performed using IBM SPSS® Statistics software (ver. 24.0; IBM Corp., Armonk, NY, USA) and SAS (version 9.4, SAS Inc., Cary, NC, USA). For the assumption of a normal distribution, we used the Shapiro–Wilk tests. According to the normality of the data, continuous variables (age, height, weight, body mass index (BMI), duration of operation, duration of anesthesia, intraoperative vital signs, EtCO2, and PaCO2) were analyzed using Student’s t test or Mann–Whitney U test and were reported as the mean±standard deviation (SD) or median [interquartile range (IQR)]. All categorical and ranking variables (sex, ASA classification, tumor location, malignancy, mortality, and incidence of POPF) were analyzed using the χ2 test or Fisher’s exact test and were expressed as n (%).

Differences in LOS and ICU stay (in days) were calculated using the Kaplan–Meier analysis and log-rank test. Perioperative factors of LOS were analyzed using stepwise Cox regression analysis. Estimates for hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated from linear regression models. Independent risk factors for LOS were studied using linear regression analysis. Eighteen variables relating to baseline status were selected for analysis. They included age, sex, operation time, ASA classification, intraoperative vital signs, comorbidities, POPF, and hypocapnia. The variables were tested for selection through stepwise variable selection and were assessed for their impact on LOS. A correlation analysis was conducted to determine the relative influence of EtCO2 and LOS. A value of p < 0.05 was considered to indicate a significant difference for all analyses.

Results

A total of 756 patients were identified, and their medical histories were reviewed. A total of 727 patients were finally included (Fig. 1). Baseline characteristics of the patients are presented in Table 1. The mean EtCO2 of the hypocapnia group was 33.3 ± 1.9 mmHg versus 36.5 ± 1.2 mmHg in the normocapnia group (p < 0.001). The mean PaCO2 in the hypocapnia group was lower than that in the normocapnia group (32.1 ± 2.6 vs. 33.4 ± 2.4, respectively; p < 0.001). Although the body weight in the normocapnia group was greater than that in the hypocapnia group, the BMI was comparable between the groups. The values of SBP, CVP, and PIP were statistically higher in the normocapnia group. There were more patients with ASA classification 3 in the normocapnia group than in the hypocapnia group.

Fig. 1
figure 1

CONSORT flow diagram of the study

Full size image
Table 1 Baseline characteristics of the patients included in the two study groups
Full size table

Primary outcome parameters

The patients in the hypocapnia group had a significantly longer hospital stay than those in the normocapnia group (median [IQR], 22 days [16–29] vs. 18 days [13–24]; p < 0.001; Fig. 2 and Table 2). The correlation between EtCO2 and LOS showed a negative correlation with r =  − 0.249, p < 0.001 (Fig. 3).

Fig. 2
figure 2

Kaplan–Meier analysis of the effect of end-tidal carbon dioxide (EtCO2) on length of hospital stay (LOS)

Full size image
Table 2 Postoperative outcomes after pylorus-preserving pancreatoduodenectomy
Full size table
Fig. 3
figure 3

Spearman's correlation analysis between end-tidal carbon dioxide (EtCO2) and length of hospital stay (LOS)

Full size image

Secondary outcome parameters

The length of ICU stay and postoperative 30-day, 1-year, and 2-year mortality did not differ between the groups (Table 2). Cox multivariable analysis revealed that hypocapnia was an independent risk factor for longer LOS (HR, 1.61; 95% CI, 1.37–1.89; p < 0.001; Table 3). POPF was also an independent risk factor for LOS (HR, 1.41; 95% CI, 1.18–1.69; p < 0.001), but the incidence of POPF did not differ between the two groups (Tables 2, 3). Age and surgeon’s experience were also significantly associated with a longer LOS.

Table 3 Perioperative factors affecting postoperative length of hospital stay
Full size table

Discussion

This study shows that patients whose EtCO2 was less than 35 mmHg during surgery had a longer hospital stay after PPPD, and this mild hypocapnia is an independent risk factor associated with a longer hospital stay (HR 1.61) when compared with patients with EtCO2 ≥ 35 mmHg.

The length of hospital stay is a valid factor related to quality of care and patients’ postoperative functioning [10]. Similar to the results of two other studies [6, 7], our investigation revealed that even mild hypocapnia can result in longer hospital stays and unwanted adverse outcomes. Because all three studies, including ours, were conducted retrospectively, we can surmise that unfavorable physiologic changes due to a low EtCO2 could result in detrimental outcomes. As mentioned earlier, hypocapnia has various effects on the whole body, most of which are negative. Hypocapnia, in particular, decreases tissue perfusion, reduces splanchnic blood flow, and affects pulmonary stress, all of which are likely related to the most common complications and factors associated with recovery. A couple of studies investigated the clinically significant effects of EtCO2 prospectively and demonstrated the harmful effects of hypocapnia in terms of postoperative nausea/vomiting or postoperative delirium [11, 12]. However, it is difficult to prove which aspects of the deteriorating pathophysiology directly affect particular organs or systems.

Meanwhile, mild hypercapnia is considered more beneficial after surgery [6, 13,14,15], while the mean EtCO2 in our normocapnic group was just 36.2 mmHg, and no patient was categorized as having hypercapnia. Our results show that hypocapnia has a correlation with negative impact on patient outcomes, and whether normocarbia or hypercapnia is more beneficial than normocapnia should be studied further.

In general, EtCO2 can be monitored simply and noninvasively, and it correlates with the arterial partial pressure of carbon dioxide (PaCO2); therefore, EtCO2 is an essential parameter to be monitored during anesthesia. Two previous studies presented and analyzed only EtCO2 values; the present study added analysis of PaCO2 values and found that these values were also significantly different between the two groups. However, unlike the gradient between the two variables, which is known to be 2–5 mmHg, it was found that the PaCO2–EtCO2 differences were negative and the hypocapnic group was more hypocarbic, while the normocapnic group had less hypocarbia. Reversal of the gradient can be observed normally in pediatric, pregnant, and obese patients, caused by low functional residual capacity (FRC) and decreased total lung compliance, mechanical ventilation with large tidal volumes and low frequency, increased cardiac output and CO2 production [16,17,18,19]. FRC is inevitably reduced in patients undergoing upper abdominal surgery under general anesthesia [20]; therefore, the trend of negative PaCO2–EtCO2 in the present study is a probable clinical manifestation. However, EtCO2 is the average value over the entire operation period, while PaCO2 is the value obtained a few moments during the operation, particularly at the moment an anesthesiologist needs to be aware of a patient’s blood gas or electrolyte status; therefore, it is not proper to predict that the negative trend of PaCO2–EtCO2 differences was maintained during the entire operation and in all patients. In accordance with previous studies, it can be concluded that EtCO2 value can predict of patients’ outcomes, and this result is clinically practical because EtCO2 is continuously monitored in all anesthetized patients, whereas PaCO2 requires an invasive technique for determination and is not continuously monitored.

This study was a retrospective study, and it is likely that patient characteristics are not comparable between the two groups. Furthermore, no information was recorded regarding the strategies of the ventilator settings. The concept of lung protective ventilation was not popular about 10–15 years ago; when comparing the annual patient distribution, the number of patients increased from hypocapnia to normocapnia over time. Anesthesiologists might have applied a familiar tidal volume and respiratory rate, rather than a cautious ventilator setting considering both height and weight. As a result, patients who were lightweight may have hyperventilated, resulting in a low EtCO2 in the hypocapnic group in the present study.

Dony and colleagues studied hospital stay as an indicator of healthcare quality as well as mortality as an indicator of the quality of critical care, and they found an increased 30-day mortality in hypocapnic patients [6]. In our study, mortality was increased for 30-day, 1-year, and 2-year periods, but there were no statistically significant differences between the normocapnic and hypocapnic groups. First, the 30-day mortality of PPPD was less than 1%, which makes it difficult to demonstrate statistical significance in a few hundred patients. Regarding the 1- and 2-year mortality, they might be affected by disease progression or severity, such as cancer stage and recurrence, rather than by acute postoperative recovery. Alternatively, more patients may be needed to reveal the effect of hypocapnia on this long-term outcome.

POPF remains the main source of major morbidity and mortality after PPPD, which can occur in 13–41% of postoperative patients [21, 22]. Patients who develop POPF are also more likely to develop postoperative complications, including prolonged LOS, wound infection, bile leak, and acute cardiac events [23]. POPF was a significant factor that also led to prolonged LOS in this study (Table 3); even so, EtCO2 did not seem to be related to POPF.

This study has several limitations. First, this is a retrospective study that reveals the relationship between phenomena and outcome, and the factors affecting LOS are so complex and diverse that it is difficult to investigate the exact mechanism of EtCO2 on LOS. Further prospective investigations are warranted to determine if there is a causal relationship. Second, because of the technical limitations of our electronic medical record system, ICU stays were measured on a daily basis rather than on an hourly basis; this may have affected the capacity to discern differences. Third, there was no analysis regarding the influence of other complications on LOS, as the incidence of complications besides POPF was minimal (< 1%) and less than expected. This may be due to insufficient medical records. The development of POPF, the major complication of PPPD, was well-documented in the medical records and was thus able to be analyzed in this study. Fourth, four surgeons were involved in the study over 10 years. Even though their surgical skills and outcomes were considered comparable and their discharge criteria were standardized, it is difficult to assume that they were perfectly identical. Finally, the LOS of PPPD can differ depending on each country’s medical system including insurance, or each institution’s policy or the application of the enhanced recovery after surgery (ERAS) protocol. Therefore, there is an issue of external validity. Further studies are needed on the effects of EtCO2 in institutions with shorter hospital stays.

In summary, intraoperative end-tidal CO2 status significantly influenced LOS in our cohort of patients undergoing PPPD. This study suggests that preventing hypocapnia during general anesthesia may shorten the length of hospital stay after major abdominal surgery such as PPPD. A prospective study is warranted to demonstrate causality and to determine the best CO2 management strategy for these and other procedures.