Epidural Anesthesia–Analgesia and Recurrence-free Survival after Lung Cancer Surgery: A Randomized Trial

This was a randomized controlled trial with two parallel arms. The study protocol was approved by the Clinical Research Ethics Committee of the Peking University First Hospital (2013[653]; principal investigator: Dr. Wang) on December 27, 2013. The latest version study protocol (V1.4) was approved on September 22, 2017 (Supplemental Digital Content 1, http://links.lww.com/ALN/C645). The study was registered a priori with the Chinese Clinical Trial Registry (www.chictr.org.cn, ChiCTR-TRC-14004136; January 2, 2014) and ClinicalTrials.gov (NCT 02801409; June 15, 2016).

This trial was initiated after the end of patient recruitment of our previous trial investigating the impact of epidural anesthesia–analgesia on postoperative delirium with a long-term follow-up.^23,24  The trial was conducted in Peking University First Hospital in Beijing, China, in accordance with Good Clinical Practice guidelines and the Declaration of Helsinki. Written informed consent was obtained from participating patients or authorized surrogates. The effect of combined epidural anesthesia on tumor-infiltrating lymphocytes in lung adenocarcinoma has been previously reported in a subset of these patients.²⁵ 

We enrolled adults aged 18 to 80 yr who were clinically diagnosed as lung cancer, were scheduled for radical surgery, and requested patient-controlled postoperative analgesia. Exclusion criteria were: (1) distant metastasis, malignant tumor in other organs, or preoperative chemotherapy, radiation therapy, or immune therapy; (2) comorbid autoimmune diseases, glucocorticoid use, or immunosuppressant therapy within 1 yr; (3) severe neurologic conditions, hepatic disease (Child–Pugh classification C), renal failure (serum creatinine greater than 442 μmol · l^–1), renal replacement therapy, or American Society of Anesthesiologists (ASA) physical status classification of IV or higher; (4) history of anesthesia and/or surgery within 1 yr; (5) contradictions to epidural anesthesia, including spinal deformity, coagulation dysfunction, local infection, and history of spinal trauma/surgery; or (6) allergy to any trial-related medication.

Random numbers were generated by a biomedical statistician using the SAS 9.2 software (SAS Institute, USA) with a block size of 4 and were concealed in sequentially numbered opaque envelops. The envelopes were opened by an investigator shortly before induction of anesthesia; allocation was thus concealed as long as was practical. During the study period, the enrolled patients were randomly allocated to receive either general anesthesia alone plus postoperative intravenous analgesia or combined epidural–general anesthesia plus postoperative epidural analgesia in a 1:1 ratio without stratification.

No premedication was given. Routine intraoperative monitoring included electrocardiogram, noninvasive blood pressure, pulse oxygen saturation, end-tidal concentrations of inhaled anesthetics and carbon dioxide, nasopharyngeal temperature, Bispectral Index, and urine output. Invasive arterial pressure and central venous pressure were monitored when necessary. In each case, surgery was video-assisted thoracoscopic tumor resection.

For patients assigned to general anesthesia alone, anesthesia was induced with propofol, sufentanil, and rocuronium, with or without midazolam. Anesthesia was maintained with propofol infusion and/or sevoflurane inhalation, with or without nitrous oxide inhalation, supplemented with opioids (remifentanil infusion and sufentanil infusion/injection) and muscle relaxants (rocuronium or cisatracurium). Dexmedetomidine was given at the discretion of anesthesiologists. The target was to maintain a Bispectral Index between 40 and 60. A double-lumen endobronchial tube or a bronchial blocker was used to facilitate one-lung ventilation. A mixture of oxygen and air/nitrous oxide was provided during two-lung ventilation and also during one-lung ventilation as long as the pulse oxygen saturation was higher than 93%. After surgery, patient-controlled intravenous analgesia was provided for up to 3 days. The analgesic infusion was morphine 0.5 mg · ml^–1, and the pump was programed to deliver 2-ml boluses with a lockout interval of 8 min and a background infusion rate of 1 ml · h^–1.

For patients assigned to combined epidural–general anesthesia, an epidural catheter was inserted before anesthetic induction. The intervertebral space was selected according to the site of incision, usually between T5 and T8. A test dose of 2% lidocaine was injected to confirm the position of epidural catheter. Epidural anesthesia was maintained with intermittent boluses of 0.375% ropivacaine until the end of surgery. General anesthesia was induced and maintained as for patients assigned to general anesthesia alone, again titrated to a Bispectral Index between 40 and 60. After surgery, patient-controlled epidural analgesia was provided for up to 3 days. The analgesic solution was a mixture of 0.12% ropivacaine and 0.5 μg · ml^–1 sufentanil. The pump was programed to deliver 2-ml boluses with a lockout interval of 20 min and a background infusion rate at 4 ml · h^–1. If an epidural catheter could not be inserted, patients were given general anesthesia and intravenous analgesia as above.

Low-dose glucocorticoids (usually 5 mg dexamethasone) were administered before anesthesia induction as prophylaxis against postoperative nausea and vomiting. Intraoperative hypotension was managed by reducing anesthetic depth, fluid infusion, and/or administration of vasopressors (ephedrine, phenylephrine, norepinephrine, dopamine, and/or epinephrine). Intraoperative bradycardia was managed with atropine and/or other chronotropic agents (dopamine, epinephrine, and/or norepinephrine).

Patients were monitored in the postanesthesia care unit for at least 30 min before being transferred back to their wards. Intensive care unit (ICU) beds were reserved for patients with preexisting conditions such as severe reduction in ventilatory and/or diffusion function, preexisting respiratory failure, or significant cardiovascular comorbidities, pneumonectomy, and advanced age.²⁶  However, their beds were released if they had smooth and uneventful anesthetics. Other patients were unexpectedly admitted to an ICU when they experienced massive intraoperative bleeding or unstable hemodynamic conditions or were difficult to extubate. Disposition decisions were made collaboratively by attending anesthesiologists and surgeons.

Patients with insufficient postoperative analgesia were initially managed with patient-controlled analgesia; supplemental analgesics including nonsteroidal anti-inflammatory drugs and other opioids were given when necessary. Per routine, pain management in the postanesthesia care unit was performed by anesthesia nurses and anesthesiologists and in the ICU by nurses and intensivists. In surgical wards, patient-controlled analgesia was guided by anesthesia nurses who evaluated patients twice daily; supplemental analgesics were prescribed by surgeons. The target was to maintain a numeric rating scale level (an 11-point scale where 0 = no pain and 10 = the worst pain) of pain at rest of 3 or lower.

Postoperative chest physiotherapy was performed routinely to improve ventilation and promote sputum clearance. Patients were encouraged to spend time out of bed within 24 h. Chest drainage tubes were removed when there was no longer an air leakage, atelectasis resolved, and drainage volume of less than 200 ml of clear fluid within 24 h.²⁷  Patients were usually discharged from the hospital a day after their chest tubes were removed.

Anesthesiologists and patients were aware of group assignment, as were the investigators assessing immediate postoperative management (S.-M.H. and H.Kong). However, all long-term outcomes including cancer recurrence and mortality were performed by a separate group of investigators (Z.-Z.X., Q.-H.L., and X.-Q. Shang) who did not participate in anesthesia/surgery and had no knowledge of trial group assignment; those investigators were forbidden to discuss type of anesthesia with patients or other healthcare providers.

Baseline data included demographic and morphometric characteristics, preoperative surgical diagnosis, medical comorbidities, previous surgery, family history of cancer, personal history (smoking, drinking, and contact with potentially carcinogenic substances), laboratory and other examinations, as well as clinical pathologic type and tumor-node-metastasis stage if available. Intraoperative data were recorded by attending anesthesiologists and included type and duration of anesthesia, type and doses of anesthetics and other medications, volume of estimated bleeding, blood transfusion, fluid balance, vital signs (heart rate, blood pressure, and pulse oxygen saturation), and arterial blood gas results, as well as type, location, and duration of surgery.

Vital signs were monitored continuously in patients admitted to an ICU. On surgical wards, vital signs were evaluated noninvasively at 15- to 30-min intervals until the first postoperative morning and then once or twice daily until hospital discharge. Pain intensity at rest and with coughing was assessed with the 11-point numeric rating scale between 8 and 10 am during the first 3 days; a minimum difference of 1 point was considered clinically meaningful.²⁸ 

We recorded analgesics including opioids contained in patient-controlled analgesia infusions and rescue analgesics. Administered opioids were converted to intravenous morphine equivalents.²⁹  Postoperative complications were defined as new-onset medical conditions that were deemed harmful and required therapeutic intervention (i.e., grade II or higher on the Clavien–Dindo classification).³⁰  Duration of chest-tube placement and length of hospital stay were recorded. The results of pathologic examinations and tumor-node-metastasis staging³¹  were documented.

Patients were advised to return to the hospital for evaluation every 3 months during the initial year, every 6 months during the second year, and yearly thereafter.³²  Patients with lymph node metastasis (tumor-node-metastasis stage II or higher) at the time of surgery were asked to return every 3 months throughout. Investigators contacted patients at the designated intervals to remind them to return and to ask about recurrence, cancer treatment, and vital status. Specific evaluations were prescribed by thoracic surgeons and usually included chest radiography, computed tomography scans, positron emission tomography scans, sputum cytology, and serum tumor markers.^32,33  Test results and clinical diagnoses were collected from our hospital’s medical information system. Occasionally, patients were unable to return for a particular visit, in which case their medical records and examination results were requested from the relevant institution and when possible and then verified on subsequent visits to our hospital.

Data collected during each postoperative patient contact included the following: (1) Whether anti-cancer therapies were given. (2) Results of interval examinations. Cancer recurrence was defined as reappearance of the same cancer in the ipsilateral thorax, including lung and/or mediastinal/hilar lymph nodes. Metastases were defined as reappearance of the same cancer in any other part of the body. Cancer recurrence and/or metastasis was diagnosed by thoracic surgeons (and/or radiologists); time of the earliest diagnosis was recorded. For patients who had palliative resections or had unresectable cancers, cancer progression was defined by the first of an increase in tumor diameter of 2 mm or more, appearance of new metastatic foci, or death. (3) Vital status, including date of death. Cancer-specific death was defined as death fully attributable to lung cancer for which surgery was performed and usually occurred after cancer recurrence/metastasis after exclusion of other causes such as stroke, myocardial infarction, and accidents.

Among patients surviving 1 yr, metabolic equivalents (1 metabolic equivalent is equal to 3.5 ml·min^–1 · kg^–1 resting oxygen consumption) during daily life activities were assessed. Full physical recovery was defined as engagement in moderately intense activity (3 to 5.9 metabolic equivalents) for at least 150 min a week.^34,35 

Our original primary outcome was cancer recurrence. However, during the follow-up period, we noted that some cancer patients died before recurrence/metastasis. Before analysis and without accessing trial data, we therefore changed the primary outcome to recurrence-free survival, defined as time from surgery to the earliest date of recurrence/metastasis or death from any cause, whichever came first. Simultaneously, we added recurrence-free, overall, and cancer-specific survival in patients with confirmed cancer. Secondary endpoints included overall survival and cancer-specific survival. Overall survival was defined as time from surgery to all-cause death. Cancer-specific survival was defined as time elapsed between surgery and cancer-specific death, with deaths from other causes being censored.

Other secondary endpoints included ICU admission, the incidence of postoperative complications, duration of chest tube placement, postoperative hospital duration, in-hospital mortality, and physical activity among 1-yr survivors. Other prespecified outcomes included pain intensity during the first 3 postoperative days. Post hoc subgroup analyses were performed as functions of age, sex, chronic smoking, ASA physical status, tumor-node-metastasis stage, and postoperative anticancer therapy.

Trial-related adverse events were evaluated from initiation of anesthesia until the 3rd postoperative day. Failed epidural catherization was identified by the responsible anesthesiologist. Intraoperative high airway pressure was defined by peak airway pressures greater than 30 cm H₂O unrelated to mechanical factors such as kinked or misplaced endotracheal tubes. Bradycardia was defined as heart rate less than 45 beats/min. Hypotension was defined as systolic blood pressure less than 90 mmHg or a decrease of more than 30% from individual preoperative ward values. Hypertension was defined as systolic blood pressure greater than 180 mmHg or an increase of more than 30% above preoperative ward values.

In a previous study of patients who had complete resections for non–small cell lung cancer, one third had tumor recurrences within a median duration of 24 months.³⁶  In a pilot investigation in our institution, the hazard for recurrence within a year after lung cancer surgery was 48% lower in patients with combined epidural–general anesthesia compared with general anesthesia alone. We anticipated that patient recruitment would take 2 yr and planned to follow patients for at least 2 yr thereafter and that the 2-yr recurrence incidence would be 33% in patients with general anesthesia alone.

With the two-sided significance level set at 0.05 and power at 80%, an estimated sample size of 360 participants (180 per group) was required to detect a one-third reduction in recurrence. Considering a dropout rate of 8% and an epidural failure rate of 2% according to our own experience, we therefore planned to enrolled 400 patients without interim analyses. Sample size was estimated by log-rank test with PASS software (version 11.0, NCSS PASS, USA). Enrollment ceased when the target sample size was reached.

Baseline balance was assessed with absolute standardized differences, which are defined as the absolute difference in means, mean ranks or proportions divided by the pooled SD and calculated using the formula published by Austin.³⁷  Baseline data with an absolute standardized difference greater than or equal to 0.196 (i.e., 1.96×√(n1 + n2)/(n1 × n2)) were considered imbalanced between groups.

Recurrence-free survival was analyzed using a Kaplan–Meier estimator with differences between groups assessed by log-rank test; a Cox proportional hazard model was used to adjust for factors predetermined according to clinical importance and included age, sex, chronic smoking, ASA physical status classification, tumor-node-metastasis stage, and postoperative anticancer therapy. Effect size was expressed as hazard ratio and 95% CI. The interactions between treatment effect and predefined factors as above were assessed separately with Cox proportional hazard models. Schoenfeld residual was used to test proportional hazard assumptions.

Secondary endpoints including overall and cancer-specific survival, as well as subgroup endpoints including recurrence-free, overall, and cancer-specific survival in cancer patients were analyzed using Kaplan–Meier estimators with differences between groups assessed by log-rank tests; the Cox proportional hazard models were used to adjust for the predetermined factors listed above.

For other outcomes, numeric variables were analyzed using independent-sample t or Mann–Whitney U tests; differences (and 95% CIs for the differences) between medians were calculated with Hodges–Lehmann estimators. Categorical variables were analyzed with the chi-square tests, continuity correction chi-square tests, or Fisher exact tests. Other time-to-event variables such as duration of chest tube drainage and duration of postoperative hospitalization were analyzed with Kaplan–Meier estimators, with differences between groups assessed with the log-rank tests. Ordinal variables such as pain scores at rest and with cough were analyzed with Mann–Whitney U test.

Outcome analyses were performed in the intent-to-treat population. A per-protocol analysis was also performed for the primary endpoint. For all hypotheses, two-tailed P values <0.05 were considered statistically significant. For interactions between treatment effect and predefined factors, P values <0.10 were considered statistically significant. Multiple testing for secondary endpoints increases the risk of type I error, but we did not correct for multiplicity.

All statistical analyses were performed with SPSS 25.0 software (IBM SPSS, USA) and statistical packages R (http://www.r-project.org; Mirrors https://mirrors.tuna.tsinghua.edu.cn/CRAN/; version 3.6.1, Austria).

Between May 25, 2015, and November 11, 2017, 400 patients were enrolled and randomly assigned to either general anesthesia alone (n = 200) or combined epidural–general anesthesia (n = 200). All were included in the intent-to-treat analysis. There were six protocol deviations (two had postoperative analgesia less than 24 h because of severe postoperative nausea and vomiting, and four changed from epidural to intravenous analgesia because of inadequate analgesia or massive bleeding), leaving 394 patients in the per-protocol analysis. One year after surgery, one patient was lost to follow-up and fourteen patients died; others completed activity assessment. At completion of the follow-up period (median, 32 months; interquartile range, 24 to 48 months), 7 patients (1.8%) were lost to follow-up (fig. 1). Follow-up ended November 30, 2019.

Among all enrolled patients, 84% (336 of 400) had histologically confirmed lung cancer. Baseline characteristics were generally comparable in the randomized groups. However, the percentage of patients with tumor-node-metastasis stage 0 and negative anaplastic lymphoma kinase gene detection were slightly lower in patients assigned to combined epidural–general anesthesia, whereas the percentage with mild ventilatory function reduction was slightly higher (table 1; tables S1 and S2 in Supplemental Digital Content 2, http://links.lww.com/ALN/C646).

As expected, patients in the combined epidural–general anesthesia group were given less opioid and had lower mean arterial pressures during surgery. They received epidural sufentanil per protocol but were given less intravenous morphine over the initial 3 postoperative days. Total perioperative opioid consumption was about 8% less in patients assigned to combined epidural–general anesthesia (table 2; table S3 in Supplemental Digital Content 2, http://links.lww.com/ALN/C646).

Only four deaths occurred before confirmed cancer recurrence. Specifically, one patient assigned to general anesthesia alone died from chemotherapy-related toxicity during the initial postoperative month; three others with combined epidural–general anesthesia died from chemotherapy-induced toxicity, severe brain injury, and acute myocardial infarction; none had evidence of cancer recurrence/metastasis, although all four had histologically confirmed lung cancer during their index operations. The two deaths caused by chemotherapy without recurrence/metastasis were judged to be cancer-related (table S4 in Supplemental Digital Content 2, http://links.lww.com/ALN/C646). In our patients, recurrence-free survival was thus equivalent to being recurrence-free.

Recurrence-free survival did not differ between the two groups (fig. 2). When follow-up ended, there were 54 events (recurrence or death) among the 200 patients (27%) randomized to general anesthesia alone, compared with 48 events in the 200 patients (24%) given combined epidural–general anesthesia during a median 32-month period (adjusted hazard ratio, 0.90; 95% CI, 0.60 to 1.35; P = 0.608). The proportional hazard assumption was not violated (P = 0.168). Per-protocol analysis also showed no differences with 54 events in 199 patients (27%) assigned to general anesthesia alone versus 47 events in 195 patients (25%) given combined epidural–general anesthesia (adjusted hazard ratio, 0.92; 95% CI, 0.61 to 1.39; P = 0.688; table 3; table S5 in Supplemental Digital Content 2, http://links.lww.com/ALN/C646). No subgroup interactions were statistically significant (fig. 3).

There were no statistically significant or clinically meaningful differences in overall and cancer-specific survival between the two groups (table 3; table S5 in Supplemental Digital Content 2, http://links.lww.com/ALN/C646). In the subgroup of patients with confirmed lung cancer, there were no significant differences in recurrence-free, overall, or cancer-specific survival between the two groups (table S6 in Supplemental Digital Content 2, http://links.lww.com/ALN/C646).

Patients assigned to combined epidural–general anesthesia were less likely to be admitted to the ICU after surgery (relative risk, 0.200; 95% CI, 0.044 to 0.901; P = 0.032) and had chest tubes removed earlier (hazard ratio, 1.23; 95% CI, 1.00 to 1.50; P = 0.019) and shorter postoperative hospitalization (hazard ratio, 1.29; 95% CI, 1.06 to 1.58; P = 0.004). Pain scores recorded on the first three postoperative mornings at rest (median difference, –1 to 0 points; all P < 0.001) and during coughing (median difference, –1 point; all P < 0.001) were significantly lower in patients with combined epidural–general anesthesia than in those with general anesthesia alone (table 4; tables S7 and S8 in Supplemental Digital Content 2, http://links.lww.com/ALN/C646).

Epidural catheter insertion failed in 3 of 200 patients (1.5%). Patients assigned to combined epidural–general anesthesia had less intraoperative high airway pressure, bradycardia, and hypertension and required less treatment for hypertension. However, they more often experienced intraoperative hypotension and were more often given vasopressors. One patient in the combined epidural–general anesthesia group developed severe ischemic–hypoxic brain injury, which was attributed to severe intraoperative hypotension caused by pulmonary artery rupture and massive bleeding that was deemed unrelated to the trial; the patient died 5 months after surgery (table 5).

Combined epidural–general anesthesia with epidural analgesia did not improve recurrence-free survival compared to general anesthesia with opioid analgesia, nor did it improve other survival outcomes in the entire population or in patients with confirmed lung cancer. The only previous robust trial evaluating the effect of regional analgesia on cancer recurrence was for breast cancer, a far less invasive procedure. The breast cancer trial enrolled more than 2,100 patients, whereas we recruited 400 lung cancer patients considering that lung cancer recurs far more often. Although we reached the target number of outcome events, the CI for recurrence-free survival ranges from a 40% reduction in the hazard to a 35% increase. It therefore remains possible that regional anesthesia and analgesia reduces lung cancer recurrence by amounts that are clinically meaningful. Nonetheless, available evidence does not support the theory that regional anesthesia–analgesia reduces cancer recurrence in humans, despite strong in vitro and animal support.

As might be expected, most deaths were cancer-specific, usually resulting from cancer recurrence or metastasis. Our recurrence rate was lower than previously reported,^36,38  possibly because tissue diagnoses are rarely available before lung cancer surgery. We thus enrolled patients believed to have potentially resectable lung cancer. In fact, 16% of our patients turned out to have noncancer diseases. When noncancer patients were excluded, 32% of our patients had a recurrence within a median of 32 months, which is similar to previously reported results (33% within a median of 24 months).³⁶ 

Epidural block is generally safe in patients without contraindications and usually successful. Epidural catheterization failed in only 3 of 200 (1.5%) patients, which was lower than previously reported results (6.1%).³⁹  Severe adverse events were rare, and none was attributed to epidural block per se. However, patients with epidural anesthesia had a 22% higher incidence of intraoperative hypotension and therefore the need for vasopressors, a well known consequence of combining epidural and general anesthesia.⁴⁰  Intraoperative hypotension is associated with adverse events including delirium,⁴¹  myocardial injury,⁴²  acute kidney injury,⁴³  and mortality.⁴⁴  Our trial was not powered for specific organ injuries, although we evaluated a composite of serious complications and mortality. We were therefore unable to fully evaluate the overall risks and benefits of intraoperative epidural anesthesia.

Unsurprisingly, epidural anesthesia required about 46% less intraoperative opioid; epidural analgesia also provided superior postoperative pain relief, with a median 1 point lower on our 0- to 10-point pain scale. Overall, epidural anesthesia–analgesia consumed about 8% less opioid, which is consistent with previous results.^45,46  The difference was smaller than might be expected because the epidural infusion included the opioid sufentanil per clinical routine.

Better analgesia and less opioid consumption likely explain other short-term benefits consequent to epidural–general anesthesia. For example, good analgesia might have improved the patients’ ability to cough and recover good lung function, thereby allowing chest tubes to be removed earlier. Similarly, 80% fewer patients given epidural analgesia required postoperative ICU admission, possibly because pain was well controlled with less opioid. Shortened hospitalization (median 1 day shorter, corresponding to a 17% decrease) may have been a natural consequence in patients randomized to epidural–general anesthesia given earlier chest tube removal⁴⁷  and fewer ICU admissions.⁴⁸ 

An important limitation is that clinical teams could not be blinded to analgesic strategy. Clinician expectations may thus have influenced the ICU admission, timing of chest tube removal, and discharge decisions. Although it seems unlikely that all apparent benefit and harm were due to clinician bias, some presumably was. Pain was assessed only once daily, starting the first postoperative morning. Furthermore, a difference of 1 point is of marginal clinical importance,²⁸  although we might have missed greater differences that are usually apparent during the initial postoperative hours.

As might be expected in a moderate-sized randomized trial, baseline balance was good. However, there were slight imbalances in tissue diagnoses and tumor extent. We therefore included cancer stage and other factors in a multivariable model for correction and performed a subgroup analysis in patients with confirmed cancer, which confirmed the primary results. With a total of 102 outcome events, we had reasonable power for identifying moderate treatment effects. However, it remains possible that regional anesthesia and analgesia reduces recurrence by amounts that might be considered clinically meaningful. Our single-center approach reduces generalizability of our findings. Nonetheless, our anesthetic approach was fairly routine and not overly proscriptive. Given our equivocal results, it seems unlikely that findings would much differ with any similar approach, even in another institution.

In summary, combined epidural anesthesia–analgesia for lung cancer surgery did not improve recurrence-free, overall, or cancer-specific survival compared to general anesthesia alone. Our trial was powered to detect a relative reduction in cancer recurrence of about a third, but was underpowered for smaller effects that might still be clinically meaningful. Combined epidural anesthesia–analgesia remains a reasonable alternative for major lung cancer, although it should not be expected to reduce the risk of lung cancer recurrence.

The authors gratefully acknowledge Xue-Qian Shang, M.D. (Department of Thoracic Surgery, Peking University First Hospital, Beijing, China), Hui-Hui Wang, M.D. (Department of Medical Imaging, Peking University First Hospital), and Ting Li, M.D., and colleagues (Department of Pathology, Peking University First Hospital) for their help with long-term follow-up, acquisition of data, and pathologic consultation.

Supported by National Key R&D Program of China (Beijing, China) grant No. 2018YFC2001800 and B. Braun Anesthesia Scientific Fund (Shanghai, China) grant No. BBDF-2016-009.

The authors declare no competing interests.

Full protocol available at: dxwang65@bjmu.edu.cn. Raw data available at: dxwang65@bjmu.edu.cn.