Non-intubated anesthesia in patients undergoing video-assisted thoracoscopic surgery: A systematic review and meta-analysis

Introduction Non-intubated anesthesia (NIA) has been proposed for video-assisted thoracoscopic surgery (VATS), although how the benefit-to-risk of NIA compares to that of intubated general anesthesia (IGA) for certain types of patients remains unclear. Therefore, the aim of the present meta-analysis was to understand whether NIA or IGA may be more beneficial for patients undergoing VATS. Methods A systematic search of Cochrane Library, Pubmed and Embase databases from 1968 to April 2019 was performed using predefined criteria. Studies comparing the effects of NIA or IGA for adult VATS patients were considered. The primary outcome measure was hospital stay. Pooled data were meta-analyzed using a random-effects model to determine the standard mean difference (SMD) with 95% confidence intervals (CI). Results and discussion Twenty-eight studies with 2929 patients were included. The median age of participants was 56.8 years (range 21.9–76.4) and 1802 (61.5%) were male. Compared to IGA, NIA was associated with shorter hospital stay (SMD -0.57 days, 95%CI -0.78 to -0.36), lower estimated cost for hospitalization (SMD -2.83 US, 95% CI -4.33 to -1.34), shorter chest tube duration (SMD -0.32 days, 95% CI -0.47 to -0.17), and shorter postoperative fasting time (SMD, -2.76 days; 95% CI -2.98 to -2.54). NIA patients showed higher levels of total lymphocytes and natural killer cells and higher T helper/T suppressor cell ratio, but lower levels of interleukin (IL)-6, IL-8 and C-reactive protein (CRP). Moreover, NIA patients showed lower levels of fibrinogen, cortisol, procalcitonin and epinephrine. Conclusions NIA enhances the recovery from VATS through attenuation of stress and inflammatory responses and stimulation of cellular immune function.

Introduction Video-assisted thoracoscopic surgery (VATS), a common diagnostic and therapeutic technology using modern video technology and high-tech equipment, has the advantages of minimal trauma and incision, reduced pain and reliable curative effect [1]. Contrary to conventional thoracotomy, VATS relies on a surveillance screen, and the operation is perform using special surgical instruments through three or four, 1.5-cm chest wall incisions [2,3]. For patients with stage Ia with tumor diameter smaller than 2 cm, VATS lobectomy can reduce the loss of lung function and improve the quality of life of patients after operation [4]. The National Comprehensive Cancer Network (NCCN) guidelines also recommend VATS lobectomy as the standard procedure for resectable lung cancer [5].
In thoracic surgery, double-lumen endobronchial intubation and pulmonary isolation technology after general anesthesia can protect the contralateral bronchus and lung tissue from contamination while fully exposing the field of operation [6]. Pulmonary sequestration using a double-lumen bronchial catheter or bronchial occlusion tube can cause such complications as throat pain, nausea, and hemoptysis [7,8]. Especially under shallow anesthesia, tracheal intubation leads to severe cough, suffocation or bronchospasm, excessive excitation of the autonomic nervous system, arrhythmias, bradycardia, ventricular premature beats, ventricular fibrillation and even cardiac arrest [9].
To reduce complications of tracheal intubation and minimize the impact of one-lung ventilation (OLV), the non-intubated (NIA) or awake VATS technique has been attempted from pleural biopsy to lobectomy [10,11]. NIA is technically feasible and safe in patients undergoing major surgery such as those undergoing lobectomy for lung cancer [12]. Whether NIA offers a good benefit-to-risk ratio for certain patient groups remains unclear [13,14], such as for patients with effective persistent hypoxemia, carbon dioxide retention, extensive pleural adhesions or severe cough. Nevertheless, NIA has the potential to improve clinical recovery and prevent stress responses, particularly in high-risk subgroups such as those with impaired pulmonary function [15][16][17]. In addition, VATS lung metastasectomy with NIA can trigger milder immunological and inflammatory responses than intubated general anesthesia (IGA) [18]. However, NIA application is limited by several factors, including the difficulty to control airway management and the need for a highly experienced anesthesiologist.
In view of the limitations and the controversy of NIA technique, we aimed to perform an updated systematic review and meta-analysis comparing NIA and conventional IGA for adult's thoracic surgery in terms of mortality, other clinical and physiological outcomes, and adverse events.

Search strategy
Systematic methods were used to identify published randomized controlled trials or retrospective case-control studies that enrolled patients who underwent any type of VATS surgery with NIA or awake VATS in comparison with IGA. There were no restrictions on publication year or country. Only articles published in English were considered. To identify all relevant studies, we manually searched MEDLINE/Pubmed (from 1946 to April 2019), Embase/OvidSP (from 1974 to April 2019), and the Cochrane Central Register of Controlled Trials (CENTRAL) (from inception in 1965 to April 2019, see S1 File). Moreover, we manually screened the reference lists of included studies and review articles in order to identify studies that were not found in the original database search.
All studies comparing NIA and IGA in adults undergoing thoracic surgery were eligible for inclusion. All combinations of VATS were included, such as lung lobectomy, metastasectomy, segmentectomy, sympathectomy, lung volume reduction surgery; and all NIA strategies, including intercostal and vagal nerve block (INB), thoracic epidural anesthesia (TEA), TEA combined with INB and the Olympus LTF-160 semi-rigid pleuroscope) and target anesthesia.
In this review, we defined lung or pleural biopsy and pleurodesis, sympathectomy, talc pleurodesis surgery, bullectomy and wedge resection as minor thoracic surgery. Moderate thoracic surgery was defined as video-assisted flexible thoracoscopic surgery (VAFTS) decortication, lung volume reduction surgery, mediastinal tumor resection and VATS for nononcological thoracic disease. Major thoracic surgery was defined as lobectomy, metastasectomy and segmentectomy.

Data extraction
Three authors (Y.M.G., R.J., and M.Y.J.) used a predefined list of terms to independently identify the studies. After excluding all duplicated studies, the titles and abstracts of the potentially relevant publications were screened. If a final decision could not be made after reading the title and abstract, the full texts was assessed. Any conflicts among the three authors regarding the selection of studies was resolved through reassessment by a fourth author (F.L.). Inter-rater κ for study inclusion was 0.68, corresponding to "good" agreement [19].
For comparing patients undergoing NIA or IGA, the primary outcome was length of hospital stay. The secondary outcomes included estimated cost for hospitalization, postoperative chest tube duration, postoperative fasting time, cellular immune function, and stress and inflammatory response.

Quality assessment
The methodological quality evaluation was performed using Review Manager (RevMan, Version 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). For assessment of the risk of bias in randomized controlled trials (RCTs), two review authors (L.F. and D.X.K.) independently evaluated the methods of random sequence generation, group allocation, blinding of participants and outcome assessment, adequacy of analyses, and completeness of reporting using previously described methods [20,21]. We did not assess three domains of bias (random sequence generation, allocation and blinding) that not applicable to retrospective or observational studies [22]. Two review authors (L.F. and D.X.K.) independently assessed the methodological quality of the these included studies in terms of bias of selection, performance, detection, and attrition.

Statistical analysis
Pooled results are shown as a summary odds ratio (OR) for dichotomous variables or pooled standard mean difference (SMD) for continuous variables, together with 95% confidence intervals (CI). These statistical estimated were determined for each parameter using the STATA/MP 13.1 Software (StataCorp LP, College Station, TX, USA).
The heterogeneity of treatment effects between studies was measured using Higgins' inconsistency test (I 2 ). I 2 values of 25%, 50%, and 75% are typically defined as low, moderate, and high heterogeneity, respectively [23]. We used Mantel-Haenszel fixed-effect's model for outcomes showing low heterogeneity. If heterogeneity >50%, the DerSimonian and Laird random-effects model was applied. Data on continuous variables that were reported as median with interquartile ranges were converted to mean±standard deviation using an online tool (http://www.comp.hkbu.edu.hk/~xwan/median2mean.html) according to the methods of estimated sample mean [24] and estimated standard deviation [25]. Data on continuous variables represented in box plots or histograms were digitized using PlotDigitizer Software 2.6.8 (Reversion October 27 2015, Sun Microsystems, Philippe Zeller, French).
The probability of publication bias was assessed using Begg's test and Egger's tests. The between-trial heterogeneity was explored by stratifying studies by the following characteristics: type of study design (RCT, retrospective study or observational study), type of surgery (minor, moderate or major), and type of NIA (INB, TEA or other anesthesia). Meta-regression was adjusted for known confounders including sample size and patient age, as well as surgery and NIA type. Results were reported as coefficients with associated 95% CIs and standard errors (SEs). The adjusted R 2 showed the relative change in heterogeneity, with a negative value suggesting that covariates predicted less heterogeneity than expected by chance. Adjusted linear correlation trends with 95% CIs were constructed for overall data as well as for subgroup analyses. Subgroup meta-regression analyses were performed for the following subsets of patients: (1) NIA group undergoing minor surgery, (2) NIA group undergoing moderate surgery, (3) NIA group undergoing major surgery, (4) IGA group undergoing minor surgery, (5) IGA group undergoing moderate surgery and (6) IGA group undergoing major surgery.

Search strategy
The latest electronic search, conducted on 1 April 2019 in MEDLINE/PubMed, CENTRAL, and Ovid Embase databases, identified 1211 citations including 366 duplicated studies. After reviewing the title and abstract of 861 studies, 53 studies were selected for further evaluation according to our inclusion and exclusion criteria. After screening the full-texts of these 53 publications, 27 studies were selected for inclusion in the meta-analysis while 26 studies were excluded, comprising 19 studies that were not comparative but reported the clinical data for NIA VATS patients without a comparison or control group, three studies that were unpublished RCTs, and four studies were published in the supplementary information of other articles. Teams of two authors independently reviewed the titles and abstracts of each publication. The data were extracted from the 27 included studies according to our predefined inclusion and exclusion criteria, provided in S1 Table. The PRISMA flowchart (Fig 1) summarizes the process of articles search and selection.
NIA patients in the subgroup of TEA (17 trials, 1297 patients; SMD -0.56 days; 95% CI -0.80 to -0.31) and INB (four trials, 221 patients; SMD -0.68 days; 95% CI -1.25 to -0.11) showed shorter postoperative hospital stay than IGA patients (Table 2). Additionally, in one study of patients undergoing bullectomy and lobectomy, hospital stay was both shorter for non-intubated patients than for intubated ones (P < 0.001 and P = 0.022); however, there were no difference in hospital stay between non-intubated or intubated patients undergoing pulmonary wedge resection.
There was no evidence of publication bias by Begg's rank correlation test (P = 0.779) or Egger's rank test (P = 0.508) (Fig 5).

Secondary outcomes
Estimated costs. A decrease in the estimated cost was found in favor of NIA compared with intubated VATS (SMD -2.83 US $, 95% CI -4.33 to -1.34; P < 0.001; nine studies with 1056 participants) (Fig 6). This finding was consistent in retrospective studies (six studies including 913  patients; SMD -3.39 US $; 95% CI -5.44 to -1.34) and in one observational study (60 patients; SMD -2.08 US $, 95% CI -2.74 to -1.41). No statistical significance difference was observed in two RCTs with 83 participants. In the analysis stratified by type of surgery, the estimated cost to VATS patients was significantly lower for NIA than for IGA participants ( Table 2).
Chest tube duration. NIA resulted in a shortened chest tube duration after surgery compared with IGA administration (SMD -0.32 days, 95% CI -0.47 to -0.17; P < 0.001; eight studies with 707 participants) (Fig 7). The difference was significant in the subgroup of clinical trials (seven trials including 647 patients; SMD -0.31 days; 95% CI -0.47 to -0.15), in studies involving major surgery (four studies with 532 patients; SMD -0.27 days; 95% CI -0.45 to -0.10) and one study involving moderate surgery (33 patients; SMD -1.56 days; 95% CI -2.37 to -0.75). However, one observational study showed no significant difference (60 patients; SMD -0.46 days; 95% CI -0.97 to 0.06). Studies involving minor surgery showed comparable chest tube duration between NIA and IGA patients (three studies with 142 patients; SMD -0.28 days; 95% CI, -0.62 to 0.06). Studies involving TEA-NIA also showed shorter chest tube duration for NIA than IGA (five studies with 544 patients; SMD -0.30 days; 95% CI -0.47 to -0.13) but this was not the case for INB (three studies including 163 patients; SMD -0.56 days; 95% CI -1.34 to 0.21) ( Table 2). Postoperative fasting time. Four studies reported on postoperative fasting time [12,16,17,31]. Pooled results proved that postoperative fasting time in patients undergoing NIA VATS was shorter than in IGA patients (four studies with 637 patients; SMD, -2.76 days; 95% CI -2.98 to -2.54) (Fig 8). Subgroup analysis in study design, type of surgery and NIA calculating SMD showed an identical trend toward shorter postoperative fasting time in the NIA group than in the IGA group (Table 2). Cellular immune function. We analyzed some variables to assess the effect of NIA on perioperative cellular immune function of thoracic patients. Interestingly, compared to IGA, NIA was associated with higher number of total lymphocytes (two studies with 118 patients; SMD, 0.32; 95% CI 0.08 to 0.56), T helper/T suppressor cell ratio (two studies, 118 patients; SMD 0.28; 95% CI 0.04 to 0.52) and number of natural killer (NK) cells (three studies with 176 patients; SMD 0.74; 95% CI 0.54 to 0.95). There were no significant differences in the levels of B lymphocytes, T lymphocytes, CD4+ T lymphocytes or CD8+ T lymphocytes between NIA and IGA groups (Table 3).

Meta-regression
Meta-regression analysis showed that SMD of hospital stay did not vary significantly with publication year, age, sample size or study type (all P > 0.05). Regarding the type of NIA, Fig 9 shows the plot of log SMD on those patients undergoing moderate surgery. The meta-regression model showed a regression coefficient b of 0.572 and a P value of 0.036 (see Fig 9). It suggested that the SMD for patients undergoing moderate surgery accounted for 68.0% source of Meta-analysis of nonintubated anesthesia in VATS inter-study heterogeneity, and removing such patients from the analysis reduced inter-study tau2 from 0.7330 to 0.2342. At the same time, meta-regression analysis confirmed that the SMD for patients undergoing minor or major surgery were not a source of inter-study heterogeneity (both P > 0.05). Meta-regression analyses of multiple covariates showed that SMD of hospital stay did not vary significantly with (1) type of study or type of NIA; (2) type of study and surgery, or (3) type of study, NIA and surgery (all P > 0.05).

Discussion
On the basis of pooled data from 27 studies that included nearly 2929 patients undergoing VATS with NIA or IGA, we found that NIA might enhance recovery through attenuation of stress and inflammatory responses and stimulation of the cellular immune function, resulting in decreased hospital stay and estimated cost of hospitalization.
OLV and general anesthesia play a pivotal role in the activation and release of stress hormones. A previous study showed that OLV induction immediately increases cortisol and epinephrine levels while simultaneously reducing NK cells activity, suggesting that OLV significantly affects metabolism and antimicrobial immune responses [47]. Since cortisol release is regulated by circulating cytokines such as IL-6 and IL-8 as well as TNF-α, OLV may upregulate cortisol plasma level by increasing production of proinflammatory factors, with the subsequent release of these soluble inflammatory mediators into the blood circulation [31,48,49]. Sympathetic excitation and increased secretion of pituitary proadrenal cortex with upregulates of epinephrine, norepinephrine and cortisol concentrations This results in elevation of mean artery pressure heart rate and blood sugar during OLV and general anesthesia, which can be interpreted as a protective response to anesthesia-related hypoxia and hypotension. Intubationinduced stress responses should not occur in NIA, while OLV and TEA can block the afferent and efferent neural pathways to inhibit the sympathetic system [50,51]. In the present study, NIA patients showed lower levels of fibrinogen, cortisol, procalcitonin and epinephrine than IGA participants, suggesting that NIA attenuates the activation and release of stress hormones.
Surgical separation, stress response, pulmonary ischemia-reperfusion and anesthetics during VATS can activate immune inflammation, such as the release of inflammatory factors and the impairment of pulmonary function after operation [52]. Compared with the IGA group, NIA patients showed significantly lower levels of white blood cells, IL-6, IL-8 and CRP whileas levels of IL-10 were similar between the two groups. This may indicate that NIA patients experienced significantly less acute infection or tissues injury and lower levels of pro-inflammatory cytokines than intubated patients under general anesthesia. Epidural anesthesia can protect against lymphocyte suppression at 72 h postoperatively, while transiently reducing NK activity through decreased epinephrine concentration [53][54][55]. An epidural catheter is not administered for NIA during VATS but it can provide effective analgesia after surgery. Here, the overall level of total lymphocytes, NK cells and T helper/T suppressor cell ratio were higher in the NIA group, while levels of T lymphocytes, B lymphocytes, CD4+ T lymphocytes and CD8+ T lymphocyteswere similar between NIA and IGA patients. This suggests that NIA can ameliorate immune suppression after surgery in some aspects.
On the other hand, some anesthesiologists have proposed that NIA increases the risk of hypoxemia, hypercapnia, mediastinal oscillation and cough reflex. Therefore, anesthesiologists should always be alert to changes in vital signs during the operation. In some cases, laryngeal mask ventilation, stellate ganglion or vagus nerve block, and preoperative inhalation of lidocaine are needed, increasing anesthesiologist's workload. In fact, NIA had proved to be a feasible and reliable technique to guarantee the stability of circulation and respiration during VATS Our results demonstrate that NIA patients realized short-term outcomes earlier than IGA patients, including shorter chest tube duration and postoperative fasting time. This accelerated recovery may help explain the lower overall estimated costs of hospitalization.
In a previous meta-analysis of patients undergoing lung resection surgery, NIA in thoracoscopic surgery favored a shorter length of hospital stay compared with IGA anesthesia, but the incidence of postoperative pulmonary complications was comparable between the two groups [13]. However, that reviews lacked a systematic search strategy, did not include an assessment of absolute changes in clinical outcomes at stress response, inflammation and cellular immune function, and did not examine perioperative outcomes. Compared with another published meta-analysis of 1684 cases [56], our study included more studies and individuals, defined hospital stay as primary outcome and provided the pooled results of estimated costs, postoperative fasting time, cellular immune function, stress and inflammatory response. This study presents several limitations. First, the present study was not registered as a systemic review and meta-analysis in PROSPERO before it was performed. Second, the results were analyzed based on study design, type of surgery and NIA level data, but not on patient level data. Third, definitions of clinical outcomes were based on the definitions in the corresponding original studies and may therefore lack uniformity across all the studies. Fourth, pooled analysis of immune function and stress responses used data from only 1-2 studies. Fifth, the majority of the included studies were retrospective and some had small samples.
In conclusion, pooled data from nearly 2929 patients receiving VATS suggest that NIA is associated with faster, better recovery from surgery than IGA, mainly reflected in shorter hospital stay and lower estimated cost. The improvement in NIA patients was evident in those undergoing moderate or major surgery, but not in those undergoing minor surgery. This improvement was associated with attenuation of stress and inflammatory response, reduced inhibition of cellular immune function. Therefore, NIA might be a safe and feasible anesthetic strategy for VATS.