The Certified Review Board of Kyoto University, Kyoto, Japan (Chairperson Prof. Shinji Kosugi) approved the protocol for this study (approval no.: R1272-3, January 23, 2020). Additionally, the informed consent requirement was waived due to this study’s retrospective nature.
Study design, setting, and population
In this single-center retrospective cohort study, we used data from the Kyoto University Hospital IMProve Anaesthesia Care and ouTcomes (Kyoto-IMPACT) database. The Kyoto-IMPACT database aims to clarify the relationship between intraoperative respiratory and cardiovascular parameters and postoperative outcomes. We consecutively selected patients who underwent surgery under the care of anesthesiologists at Kyoto University Hospital (1,121 beds). We have published several studies using the Kyoto-IMPACT database20,21. We included adult female patients aged 18 years or older who underwent gynecologic laparoscopic surgery (i.e., adnexal surgery and/or hysterectomy) at Kyoto University Hospital between January 2012 and December 2017. The gynecologic laparoscopic surgery population was selected because the predicted incidence rate of PONV in this population is 30%–40%, assumed to be a medium risk of PONV4. The exclusion criteria were as follows: 1) patients with postoperative intensive care unit admission; 2) those who underwent multiple surgeries within one week during the study period; 3) those who received epidural anesthesia; 4) those with missing smoking data, and 5) those with missing intraoperative EtCO2 data.
We collected data from the anesthesia information management and electronic medical record systems and constructed the Kyoto-IMPACT database. EtCO2 was continuously measured using a sidestream gas analyzer (GF-220R Multigas/Flow Unit, Nihon Kohden®, Japan) that was automatically uploaded to the anesthesia information management system every 60 s. Intraoperative EtCO2 was the mean EtCO2 level from skin incision to skin closure. We removed EtCO2 levels lower than 20 mmHg as artifacts (EtCO2 during aspiration or position change). The definitions of variables, including the minimum and maximum EtCO2 levels, can be found in Supplementary Data Table S1. We collected data on PONV by reviewing all clinical data contained in the electronic medical records. Ward nurses assessed the presence of nausea and vomiting at least twice daily. We defined PONV as one or more episodes of nausea or vomiting during the first two days after surgery and vomiting as one or more episodes of vomiting during the same period.
To determine how EtCO2 affects PONV, we defined exposure by calculating the dose, time, and cumulative effects of EtCO2. First, we evaluated the dose effects of EtCO2 using the mean EtCO2. Next, we divided the patients into two groups based on the cutoff EtCO2level of 35 mmHg proposed by Way and Hill22. We defined low EtCO2 as a mean EtCO2 lower than 35 mmHg and normal EtCO2 as a mean EtCO2 greater than or equal to 35 mmHg. We classified the patients in either of these groups and used them as the primary exposure for further analysis. Additionally, we categorized the mean EtCO2 levels into quartiles (i.e., <35, 35–37, 37–40, and ≥40 mmHg) because the relationship between EtCO2 and PONV might not be linear. To assess the effects of the duration and severity of low EtCO2 exposure, we determined the time effect based on the minutes when the EtCO2 level was below 35 mmHg and measured the cumulative effect as the area with EtCO2 levels below the threshold of 35 mmHg for each patient. Furthermore, we categorized the minutes and area under the threshold of an EtCO2 level of 35 mmHg into quartiles; the lowest quartile was the reference category.
The primary outcome in this study was PONV, defined as PONV for two days postoperatively. The secondary outcomes were nausea for two days postoperatively, vomiting for two days postoperatively, PONV for 3–7 days postoperatively, and PLOS. We defined PLOS as the duration of hospital stay after surgery for patients who survived until discharge.
We analyzed the relationship between intraoperative EtCO2 and PONV before data collection. We used the Mann–Whitney test for group comparisons, and continuous variables were expressed as the median and interquartile range (IQR), and categorical variables were expressed as counts and percentages (%).
First, we performed modified Poisson regression analysis with robust variance to calculate the risk ratio for low EtCO2 (mean EtCO2 of less than 35 mmHg) and PONV, with the reference category of normal EtCO2 (mean EtCO2 ≥ 35 mmHg)23. Additionally, we calculated the risk ratios of the mean EtCO2 level in the first quartile (mean EtCO2 of less than 35 mmHg), third quartile (mean EtCO2 of 37–40 mmHg), and fourth quartile (mean EtCO2 of more than or equal to 40 mmHg). The second quartile (mean EtCO2 of 35–37 mmHg) was the reference category because it was considered normocapnia. Furthermore, we examined the time and cumulative effects of EtCO2 by evaluating how each quartile affected PONV, with the first quartile (with minutes under an EtCO2 of 35 mmHg and the area below the threshold of 35 mmHg) being the reference category. We created a model using the covariates previously used to demonstrate the relationship between intraoperative EtCO2 and PONV. The model included age, smoking history, surgery duration, body mass index (BMI), total intravenous anesthesia (TIVA), mean arterial pressure (MAP), intraoperative fentanyl use, postoperative fentanyl dose for intravenous patient-controlled analgesia (IVPCA), the use of prophylactic antiemetics, the addition of droperidol to postoperative IVPCA, American Society of Anesthesiologists Physical Status (ASAPS), malignancy, and emergency surgery. Additionally, a modified Poisson regression model investigated whether the dose, time, or cumulative effect of EtCO2 affects postoperative nausea two days, vomiting two days, and PONV 3–7 days postoperatively, adjusting for the aforementioned models. To further evaluate the relationship between EtCO2 and PLOS, we performed a linear regression analysis adjusting for the possible covariates in the aforementioned models.
The relationship between intraoperative EtCO2 and PONV may depend on patient and surgical characteristics. Therefore, we performed a subgroup analysis to assess this potential heterogeneity. We used the modified Poisson regression model for the following subgroups: (i) age (≥50/<50 years), (ii) malignancy (yes/no), (iii) smoking history (ever smoked/never smoked), (iv) duration of surgery (≥4/<4 hours), (v) TIVA (yes/no), and (vi) the use of intraoperative prophylactic antiemetics (yes/no). We calculated the crude risk ratio of PONV in each subgroup and examined the interaction between subgroups and intraoperative EtCO2.
To maximize statistical power, all eligible patients enrolled in the Kyoto-IMPACT database since 2012, when postoperative nausea and vomiting began to be recorded in their current form, were included in the analysis. To determine the study power, we estimated that approximately 120 laparoscopic gynecologic surgeries were performed annually at Kyoto University Hospital, with 720 surgeries performed over six years. The risk ratio was 1.53, the incidence of PONV was 40%4, and the proportion of low EtCO2 was 50%24, giving an estimated power of 80%. The rate of missing data was 0.04%, so we conducted a complete case analysis. All statistical tests were two-tailed. We used Stata/SE 15.1 (StataCorp LLC, College Station, Texas, USA) to conduct all statistical analyses.