Survival benefits of propofol-based versus inhalational anesthesia in non-metastatic breast cancer patients: a comprehensive meta-analysis

Whether the anesthesia technique, inhalational general anesthesia (IGA) or propofol-based anesthesia (PBA), influences the long-term survival of non-metastatic breast cancer (eBC) remain unclear and controversial. We carried out a literature search on 16thJuly, 2022 for studies comparing IGA and PBA in eBC undergoing standard surgery, according to PRISMA 2020. The major endpoint in our study was overall survival (OS). Seventeen studies including four randomized clinical trials and thirteen retrospective cohort studies were included in the meta-analysis. Ten studies provided data for crude OS in unweighted eBC patients (imbalance in baseline characteristics). The summarized estimate HRs of the PBA group versus the IGA group (ten studies, N = 127,774, IGA group: 92,592, PBA group: 35,182.) was 0.83 (95%CI: 0.78–0.89). Compared with IGA, PBA was associated with both better 1-year OS (two studies, N = 104,083, IGA group: 84,074, PBA group: 20,009. Pooled HR = 0.80, 0.73–0.89) and 5-year OS (six studies, N = 121,580, IGA group: 89,472, PBA group: 32,108. HR = 0.80, 0.74–0.87). Ten studies applied PSM method to balance the baseline characteristics. In these weighted patients, PBA still showed a better OS (ten studies, N = 105,459, IGA group: 79,095, PBA group: 26,364. HR = 0.93, 0.87–1.00), a better 1-year OS (two studies, N = 83,007, IGA group: 67,609, PBA group: 15,398. HR = 0.88, 0.78–0.98) and a trend towards a better 5-year OS (nine studies, N = 121,580, IGA group: 76,797, PBA group: 24,066. HR = 0.95, 0.88–1.03). Loco-regional recurrence-free survival (LRRFS) was also better in PBA group (HR = 0.73, 0.61–0.86). The present study is the first comprehensive meta-analysis to demonstrate that propofol-based anesthesia could significantly improve OS and LRRFS in non-metastatic breast cancer patients, compared with inhalational anesthesia.


Inclusion and exclusion criteria
The inclusion criteria for eligible trials were as follows: (1) prospective/retrospective studies recruiting early-stage invasive BC patients (stage I-III); (2) patients should have undergone standard breast surgery, either mastectomy or breast conserving surgery under anesthesia; (3) studies should have included comparisons of long-term survival between IGA and PBA group; (4) detailed survival data, including overall survival (OS), recurrence-free survival (RFS), loco-regional recurrence-free survival (LRRFS) and distant metastasis-free survival (DMFS), etc., should have been presented in the studies.Definitions of survival data has been described in these studies; (4) In the case of studies derived from duplicate data sources, only the latest and the most comprehensive publication was included.
Studies meeting all the above criteria were included; otherwise, they were excluded.

Data extraction and study objectives
Two independent investigators (YJZ and LB) performed the data extraction.The following information was extracted if available: author name, publication year, journal name, type of study, data source, enrolment period and clinical stage of patients, age range/median age, race, breast surgery type (mastectomy or breast conservation), number (No.) of patients in each group, detailed anesthetic techniques and drugs, median follow-up years, balanced or not in baseline characteristics of the two groups (with weighting methods).Survival data were extracted if available,otherwise, we extracted and transformed them from the survival curve.Hazard ratios (HRs) and the corresponding 95% confidence intervals (95% CIs) of three studies (RCTs as Yan 2018, Yan 2019 and Cho 2017) could also be found in the three previously published meta-analysis [21][22][23] .The primary objective of the present meta-analysis was to evaluate the survival benefits of propofol-based anesthesia in eBC patients, compared with inhalational general anesthesia.The major index was OS, RFS, LRRFS and DMFS.

Statistical analysis
HRs and the corresponding 95%CIs for PBA v.s.IGA on survival rate were extracted or calculated from each publication.Pooled HRs and 95% CIs for endpoints presented in forest plots were estimated using the Mantel-Haenszel method, using either fixed-or random-effects models.Statistical heterogeneity was evaluated using I 2 statistics.When between-estimate heterogeneity was indicated (I 2 > 25%), a random-effects model was applied.
The quality of the eligible cohort studies was assessed using the Newcastle-Ottawa Scale (NOS).The results demonstrated that the quality of all the included studies was generally high (score 8-9).
We also applied funnel plots and Egger's test (indicated by p < 0.05) to determine the publication bias.All statistical analysis was performed with Stata 12.0 software.All tests were two-sided, and statistical significance was set at P < 0.05.
As shown in Table 1, two methods of propofol-based anesthesia (PBA) were reported in these publications: propofol-based total intravenous anesthesia (TIVA), and paravertebral blocks (PVB) with propofol sedation.For inhalational general anesthesia (IGA), most studies used sevoflurane as the inhalation anesthetics, while only one study used desflurane.
To note, Dr. www.nature.com/scientificreports/cancer patients were different in these three publications.Dr. Enlund et.al.also published three articles in Swedish population, but the inclusion period for patients and the derived database were not duplicated in the three articles.Thus we included all these publications in analysis.
As listed in Tables 2-4, HRs and their corresponding 95%CI for PBA versus IGA have been provided in most articles, while these information in three RCTs (Yan 2018, Yan 2019 and Cho 2017) with small sample size could be extracted from survival curve or from previous meta-analysis 20,21 .

Crude overall survival in unweighted patients
We treated the publications with unbalanced baseline characteristics (especially for T, N, M stage and ER, PR, HER-2 expressions) as unweighted eBC patients.The available overall survival (OS) data in such publications were curde OS.Ten studies provided data for crude OS in unweighted eBC patients (Table 2).Among them, three publications were prospective RCTs (Yan 2018, Yan 2019 and Cho 2017), but the recruited patients were unbalanced in baseline N stage and molecular subtypes.For crude OS, the separate and pooled HRs and 95% CIs were shown in Fig. 2A.No between-study heterogeneity was noted (p = 0.141, I-square = 33.4%).The summarized estimate HR of the PBA group versus the IGA group (ten studies, N = 127774, IGA group: 92592, PBA group: 35182.) was 0.83 (95% CI: 0.78-0.89).
However, the crude OS data provided in these ten publications had bias due to the different median followuptime: some were one-year OS data, while some were 5-year.To decrease the bias, we classified the data and conducted subgroup analysis.The summarized HRs for 1-year and 5-year OS were shown in Fig. 2B,C.Compared with IGA, PBA was associated with both better 1-year OS (two studies, N = 104083, IGA group: 84074, PBA group: 20009.Pooled HR = 0.80, 0.73-0.89)and 5-year OS (six studies, N = 121580, IGA group: 89472, PBA group: 32108.HR = 0.80, 0.74-0.87).The 2-year OS was similar in three publications but with a small sample size (three studies, N = 210, IGA group: 103, PBA group: 107.Pooled HR = 1.12, 95% CI: 0.90-1.40,forrest plot not shown).

Overall survival in weighted patients
Although the above results demonstrated at the first time that PBA had a better OS than IGA in eBC patients, the limitation was obvious: the baseline characteristics in the two groups were extremely unbalanced.Thus, we could not apply the results in clinical practice directly.Fortunately, we figured out that many studies had also recognized it and used propensity score matching (PSM) method to decrease the bias.We treated these OS data derived from well-balanced patients after PSM as weighted OS.
Ten studies provided weighted OS data in these well-matched eBC patients (Table 3).They were all retrospective cohort studies.For these weighted OS, the separate and pooled HRs and 95%CIs were shown in Fig. 3A.No between-study heterogeneity was noted (p = 0.946, I-square = 0.00%).The summarized estimate HR of the PBA group versus IGA group (ten studies, N = 105459, IGA group: 79095, PBA group: 26364.) was 0.93 (95% CI: 0.87-1.00).
We further classified the data according to the follow-up time and conducted subgroup analysis.The summarized HR for 1-year and 5-year OS was shown in Fig. 3B,C.Compared with IGA, PBA was associated with a better1-year OS (two studies, N = 83007, IGA group: 67609, PBA group: 15398.Pooled HR = 0.88, 0.78-0.98)

Other survival data
We also conducted analysis in other survival data, including RFS, LRRFS and DMFS (Tables 3,4).

Publication bias
Egger's test and funnel plots were used to detect and describe publication bias.No significant publication bias was identified in the data pooling (data not shown).

Discussion
Whether the anesthesia technique and drugs influences the long-term outcomes of early stage breast cancer has been always an interesting question for oncologists and anesthesiologists.Propofol has been proved to be more favorable in breast cancer surgery than volatile general anesthetics (sevoflurane) in biomarker studies: a better inhibition on VEGF-C and TGF-β, more activated natural killer cells and higher cancer cell apoptosis.
However, the differences in long-term outcomes between propofol and volatile general anesthetics were not so apparent.The previous studies were few and they reported various survival results.The most important thing to note, most studies recruited limited and unbalanced patients with various follow-up time.Thus the bias in the these studies were inherent and concrete.Several meta-analysis have been published concentrating on this question, but due to the limited sample size and unweighted data, they were not convincing to end this argument.
cancer.Thus the number of eligible studies were very limited (4 included studies).Publication of Pang QY, et.al.concentrate on the side effects of anesthetic technique, thus the number of eligible studies on survival was also very limited (3 included studies).Only Lv R, et.al.include 7 studies.Moreover, the most important issue of all the four studies is, the baseline characteristics in the include studies are unbalanced, which brings the major bias.
In the present meta-analysis, we included many latest studies with relative large sample sizes (17 studies).More importantly, many studies applied propensity score matching methods (PSM) to balance the baseline characteristics between propofol group and inhalation anesthetics group.PSM is a method to minimize selection bias between interventional groups when estimating causal intervention effects in uneven population.The treatment groups (propofol-or inhalation-anesthesia) were matched on a propensity score.The propensity score is the probability of intervention assignment conditional on the current baseline characteristics.For breast cancer patients, we treated the TNM stage and molecule subtypes (ER, PR, HER-2 expressions) as the most important characteristics.Patients should be balanced in these index after PSM.Further, we have conducted analysis according to the crude or weighted data.Thus the bias in our study has been kept to an absolute minimum, when compared with previous studies.
The most important message of present meta-analysis is that propofol-based anesthesia is associated with a better overall survival in early-stage breast cancer patients, when compared with inhalation-maintained anesthesia.It has been a long-expected but unproven clinical answer so far.The benefits could be found in both unweighted and weighted patients.Thus we think that these results could bring important impacts to the anesthesia choice for breast cancer surgery: propofol-based anesthesia should be preferred over inhalation anesthesia.We are also looking forward to the results of more well-designed undergoing RCTs, including the "CAN Study" 41 , to further confirm this question.
However, we should recognize that anesthesia period during surgery is short and once-only use for most breast cancer patients, and thus its function is hard to present.We also acknowledge that most anesthesia today are mixed, with both propofol and volatile anesthetics applied.In our cancer center, propofol are popularly used in the anesthesia induction, while volatile anesthetics are used for anesthesia maintaining.Thus, a non-propofol or a non-inhalation anesthesia are not popular nowadays, which could bring difficulties in analyze the clinical outcomes between propofol and volatile anesthetics.
Another important influencing factor is the status of eBC patients receiving surgery: whether they are newly diagnosed, or treated after neoadjuvant therapies (chemotherapy, endocrine therapy, anti-HER2 therapy, radiotherapy).In theory, the benefits of propofol mainly contribute to its function on the immune system and disseminated tumor cells during surgery.Then its efficacy may be greatly weakened in those patients with a complete or good partial remission after neoadjuvant therapies.In recent years, the ratio of neoadjuvant therapies has been increased significantly in eBC patients, which would also influence whether the patients benefit from the anesthesia technology during surgery.
After all, if we decided to apply the propofol-based anesthesia, then the next question is which is the optimal technique: propofol-based total intravenous anesthesia or paravertebral blocks with propofol sedation.In the two technology, the propofol dose and drug concentration in blood are different.More clinical trials could be conducted to further elucidate which propofol-based anesthesia could be better.

Conclusions
The present study is the first comprehensive meta-analysis to demonstrate that propofol-based anesthesia could significantly improve OS and LRRFS in non-metastatic breast cancer patients, compared with inhalational anesthesia.More prospective RCTs were needed to further clarify the target population of PBA and the optimal anesthesia usage for propofol.

for breast Figure 1. The PRISMA flow diagram.
Zhang et.al.published three articles and the patients were derived from Taiwan Cancer Registry Database (TCRD) during 2009/01/01 to 2018/12/31.However, the surgery method or PBA method

Table 1 .
Basic information of eligible studies.TIVA total intravenous anesthesia, PVB paravertebral blocks, RCT randomized clinical trials, NOS Newcastle-Ottawa Scale, n.a.not available, ref References.

Table 3 .
. and Yap A, et.al.included several kinds of cancers and did not concentrate on anesthetic technique and and prognosis breast Sum-up of weighted OS&RFS results in BC patients after PSM.