Regional anesthesia did not improve postoperative long-term survival of tumor patients: a systematic review and meta-analysis of randomized controlled trials

We searched PubMed, Cochrane library, and Embase from inception to August 4, 2021 based on PICOS (Patients: patients undergoing tumor procedures usually provided under general and regional anesthesia; Interventions: Any type of anesthesia which is implemented during tumor procedures; Comparators: eligible control groups were general anesthesia and regional anesthesia; Outcomes: the included trials reported on at least one of the following outcomes: recurrence, metastasis, survival, prognosis, death or mortality; Study design: only randomized controlled trials were included). The full electronic search strategy based on PICOS is presented in Additional file 2. We also checked the reference lists of relevant meta-analysis and reviews for additional studies.

The predefined inclusion criteria were RCTs in any language with patients undergoing any oncology surgical procedure. We compared any form of RA initiated perioperatively with any form of non-RA (GA, GA plus postoperative opioids analgesia or PCA). The primary endpoint were RFS and OS. The secondary endpoints were time to tumor progression, 5-year RFS, and 5-year OS. RFS was defined as in cancer, the length of time after primary treatment for a cancer ends that the patient survives without any signs or symptoms of that cancer; OS was defined as the length of time from the date of treatment for cancer to the death date or the censored date.

Two authors (Li and Meng) independently reviewed the studies and extracted data according the predesigned standard. Divergences of views were resolved through authors’ discussion or referring to a senior investigator (Dai). We extracted details of characteristics of the included studies. Hazard ratio (HR) was extracted and time-to-event data was also obtained [16]. Risk estimates with 95%CI value were extracted for analysis. HR was extracted from most of the included studies except “Finn 2017” and “Pi 2019”, of which only relative risk (RR) could be extracted. And RR from the two studies was pooled with HR, which is used as the statistics in the majority of the included studies.

The primary endpoint of the meta-analysis were RFS and OS. The secondary endpoints were time to tumor progression, 5-year RFS, and 5-year OS.

Methodological quality of the RCTs was independently assessed by two authors (Dai and Wang) according to the PRISMA recommendations [17]. The tool comprises seven sections: random sequence generation; allocation concealment; blinding of participants and personnel; blinding of outcome assessment; incomplete outcome data; and other bias. A score of “yes” (low risk of bias), “no” (high risk of bias), or “unclear” (risk of bias is unclear from the article) was assigned for each section.

The Mantel-Haenszel method with the fixed effects model were used to combine the dichotomous data, and the random variance method with the fixed effects model were used to combine the continuous data. Heterogeneity between studies was assessed via χ²Q statistics and I². P < 0.05 or I² > 25% was considered as existence of heterogeneity [18]. When there was significant heterogeneity among the studies (P > 0.05, I² > 50%), the random effects model was applied. On the contrary, the fixed effects model was used when homogeneous outcomes were obtained (P < 0.05, I² < 50%).

In the meta-analyses of the outcomes, pre-planned sensitivity analysis by removing one study each time, subgroup analysis and meta-regression analysis was used to investigate the origin of the heterogeneity. The sensitivity analysis shows whether the pooled estimate is stable or not. Compared with the primary I² value, a smaller I² value contributes to explain the origin of the heterogeneity. For meta-regression analysis, the magnitude of the value represents the strength of the interpretable heterogeneity and significant statistic difference was defined as a P value < 0.05. Funnel plot of each trial's effect size against the standard error was conducted to assess the possibility of publication bias. The presence of a publication bias was examined using funnel plots and the Egger regression test. Significant publication bias was defined as a P value < 0.1. The trial sequency analysis (TSA) was performed using Stata 14 with the implementation of the R 3.4.3 software, and by installing the ldbounds and metacumbounds commands. And meta-analysis was conducted using R 3.43.

We identified 905 titles and abstracts, of which 464 articles were included after screening based on title and abstract. Followed a full-text review, 15 RCTs were finally included for analyses [19‐33]. Exclusion criteria was lack of RFS or OS, not comparing RA and non-RA and plexus block using alcohol. Moreover, one study were excluded owing to unbefitting control groups. One study was published as abstract and lack of details on study variables, which was excluded (Fig. 1).

Characteristics of the selected studies are shown in Table 1. We included 15 RCTs and “Li 2022” was a secondary analysis of “Sessler 2019” which was not suitable for meta-analysis. RA technologies were epidural anesthesia (EDA), and nerve block. Regional analgesia was maintained from postoperative 48h to multiple days. The included studies yielded 5981 participants for analysis, of which 3007 participants received RA. Two study involved men only, 3 study involved women only (Table 1).

Quality of the included RCTs were presented in Fig. 2. The methodological bias of the included studies was relatively low, indicating a high qualities of the studies.

The results of our systematic review and meta-analysis indicated that RA had no significant effect on postoperative RFS and OS. This meta-analysis included 15 RCTs with 5981 participants. Heterogeneity analysis indicated no statistical heterogeneity of RA on RFS and a moderate amount of statistical heterogeneity on OS. We also established several clinically relevant subgroups to find different results. The subgroup analysis based on study design, patient characteristics and tumor type showed no statistic significant of RA on long-term survivals. Meta-regression was applied to investigate the origin of the moderate amount of statistical heterogeneity (I² = 64%) on OS. As shown in Table 2, the results of meta-regression did not find the origin of heterogeneity. So, we implemented sensitivity analysis subsequently. The sensitivity analysis showed a relevant stable result and the pooled estimate and 95%CI for OS changed significantly followed by 1 study (Binczak 2013) being removed (I² = 0%). As the heterogeneity is moderate and the sample size of the study is relatively small, we believe that removing this study has little effect on the result. Followed by “Finn 2017” and “Pi 2019” (only reported RRs) being removed, the pooled estimates of RFS and OS were − 0.03 (− 0.13 to 0.07; I², 0%) and − 0.09 (− 0.36 to 0.17; I2, 65%). There is no statistical difference between the two groups on RFS and OS followed by the two studies eliminated. I² (0% and 64%) was also similar to before (0% and 65%). So “Finn 2017” and “Pi 2019” does not pose enough interference to HRs and I² and we believe the results of the meta-analysis was stable.

Classical theories of recurrence and progression suggest that immune function is a key factor in tumor cells survival [35, 36]. Some research investigated the association between analgesia agents and tumor progression. For example, μ-opioid receptor agonist increases tumor cell migration, growth, and metastasis [37]. In contrast, local anesthetics not only inhibits tumor cell migration pathways [38]; prevents cell differentiation or proliferation [39]; and alleviates the chronic pain and anti-inflammatory effects [40, 41], but also reduces perioperative opioids. Some clinical trials showed that RA significantly improved recurrence and slowed down the progression [42, 43], resulting in prolonged survivals [44, 45] and previous meta-analysis find a positive association between RA and OS [46]. However, the above studies were research in vitro or observational studies. Some of the previous meta-analysis only investigated one type of tumor and did not include RCTs or extracted limited time-to-event data from the article [47]. This meta-analysis included all RCTs or secondary data from previous RCTs to product accurate results and showed RA could not improve cancer RFS and OS, which were not consistent with results of previous research. We predict the following reasons for this result: (1) there are many factors influencing the recurrence and metastasis of tumor. Anesthetic agents is one factor of those in vitro. But in vivo, as the internal environment is complicated, experiment in vitro is unable to reflect the situation in the human body. (2) Observation studies are lack of randomization, blinding and with confounding factors which influences the authenticity of results.

The meta-analysis has similar findings compared to the previous systematic reviews and meta-analysis but we added new RCTs with large samples of recent years (from 2019 to 2022). So, it was decided to perform the TSA for the evaluation of the analytical power of the data and assuming an RR of 15%, the 13 included studies for RFS and 12 included studies for OS provide results with adequate statistical power.

Firstly, there are 5 RCTs were purpose-designed with large sample size. The other studies included have fewer than 100 patients each group and probably contribute noise. However, risk of bias evaluation showed that the risk of bias between the studies was relative low. Meanwhile, the publication bias test showed no publication bias between the included studies. So including as many RCTs as possible contributes to maximize the sample size and reduce the confidence interval. TSA make the results more convincing.

Secondly, we analyzed several primary and secondary outcomes, which could potentially lead to an increased overall type I error rate for all outcomes under investigation.

Thirdly, we did not extend our search to conference proceedings, grey literature, and clinical trial registries. We put our best efforts into a comprehensive database search and accurate study selection strategy to provide the reader with the most updated and complete systematic review of the available literature on this topic. Embase includes conference abstracts from important medical conferences, yet, we could not identify any conference proceeding that was not eventually published in a full article form.

Fourthly, the heterogeneity in different anesthesia protocols. The scheme of general anesthesia could not be unified between the studies. Although the included studies was all RCTs with a relative high quality, combinations of different drugs and application of postoperative analgesia may mask the impact of local anesthetics.

A few RCTs with large sample and high quality are being investigated to prove the above results. NCT00684229 is a multi-center RCT which included 2500 participants to investigate the effect of RA on colorectal cancer recurrence; NCT01179308 is a multi-center RCT which included 1532 participants to investigate the effect of RA on lung cancer recurrence and NCT01975064 is a multi-center RCT which included 8000 participants to compare different general anesthesia with or without additional RA on breast, colon, or rectal cancer recurrence. Considering that most of the included studies were with small sample size and were single-center RCTs, we believe the pooled estimates will be relatively stable over time.