Initial Surgical versus Non-Surgical treatments for advanced hypopharyngeal cancer: a meta-analysis with trial sequential analysis

Background There is no consensus whether either initial surgical or non-surgical treatments should be the standard treatment pattern to advanced hypopharyngeal cancer. The aim of the meta-analysis was systematically and quantitatively to compare the relative e�cacy between initial surgical and non-surgical therapies in patients with advanced-stage hypopharyngeal carcinoma.


Background
As an aggressive malignancy, hypopharyngeal cancer is less popularity than other head and neck cancers (HNC), on behalf of 3-5% of all HNC 1 .
According to the World Health Organization (WHO) de nition, hypopharynx included four sites, namely the pyriform sinus, the post-cricoid region, the posterior wall of the hypopharynx, and the hypopharyngeal aspect of the aryepiglottic fold 2 .Usually, these patients at diagnosis present advanced stage, due to the lack of obvious symptoms in early-stage disease and prone to extensive in ltration and growth in submucosa.In addition, more than 60% patients occurred nodal metastasis in ipsilateral 3 and up to 40% of them appeared contralateral occult nodal tumor deposits 4 on presentation, which rendered formidable therapeutic challenges to ablative and reconstructive surgeons as well as clinical oncologists.The estimated 5-year overall survival rate for treated stage III and IV hypopharyngeal cancer patients varies between approximately 15% and 40%, depending on tumor-related factors, patient-related factors, and treatment approaches 5 .
Retrospect to the 1990s, curative surgical resection followed by postoperative radiotherapy was recognized as the cornerstone treatment for advanced hypopharyngeal cancer.Over the past decade, with the advancement of radiotherapy and chemotherapy, especially the application of concurrent chemoradiotherapy and targeted drugs, organ preservation and the local control rate have been remarkably improved 6 .A vast of trials, both retrospective and prospective, have revealed that organ preservation using de nitive chemoradiotherapy with surgery reserved for salvage is a receivable alternative to surgery with similar survival outcomes in HNC, including hypopharyngeal carcinoma that merely account for a smaller subgroup 7 − 9 .
Regarding the preferred therapeutic option for advanced hypopharyngeal carcinoma, there is no level-one evidence even if the adoption of chemoradiotherapy as a standard of care 10 .The determination to treat continues to be disputable in a locally advanced setting.To resolve these knowledge gaps, we performed a comprehensive meta-analysis with trial sequential analysis (TSA), which better control type I and type II errors, to compare the outcomes of surgical management (primary surgery with either adjunctive radiotherapy or adjunctive chemoradiation) versus non-surgical modalities, including induction chemotherapy with (chemo) radiotherapy and de nitive chemoradiotherapy, in patients with advanced hypopharyngeal cancer.

Methods
This meta-analysis was performed in accordance with Cochrane Handbook for Systematic Reviews of Interventions 35 and PRISMA (the Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines 36 .A pre-speci ed protocol registered at PROSPERO online platform in December 2018 (registry number: CRD42018118563).

Data sources
A systematic search of electronic databases was performed, including Medline (via the PUBMED interface), ISI Web of Knowledge, the Cochrane Library (via the Ovid interface), and EMBASE.All searches were conducted from inception through to April 10, 2019.All of the pertinent publications was identi ed by using the following terms and their combinations: (hypopharyngeal or pyriform sinus or post-cricoid), (neoplasm*or tumor* or cancer* or carcinoma* or malignanc*), (surgery or surgical) and (chemotherapy or CRT or chemoradiation or radiotherapy or RT).No language restriction was enforced.Additionally, we manually scanned references of relevant reviews of the literature and retrieved studies.To identify unpublished articles, abstracts from international conferences were inspected for additional pertinent researches.

Eligibility criteria
Literatures was independently examined by two authors to determine their eligibility for inclusion and exclusion criterion in the meta-analysis, with disagreement nally resolved by consensus.Included literatures satis ed the following PICOS standard (participants, interventions, comparators, outcomes, and study design): 1) Population: adult advanced hypopharyngeal cancer patients; 2) Intervention: surgical management (primary surgical followed by either adjuvant radiotherapy or adjuvant chemoradiation); 3) Comparison intervention: non-surgical modalities(induction chemotherapy followed by (chemo)radiotherapy and de nitive chemoradiotherapy); 4) Outcome: outcomes of interest (primary or secondary endpoint ) of the current meta-analysis; 5) Study design: retrospective cohorts or prospective cohorts or randomized clinical trial (RCT).Studies were excluded if they were case reports, letters to the editor, editorials, case series, review articles, and trials with insu cient data (e.g.lack of survival information), even if contacting the corresponding authors.

Data extraction
The relevant data was independently abstracted by two reviewers from the eligible studies into standardised collection tables for the evidence and outcomes, such as lead author, year of publication, nation, study design, number and percentages of man, age of patients, simple size, percentage of T4 stage, percentage of positive lymph node, percentage of stageIV, percentage of pyriform sinus, follow-up time, treatment modalities, overall survival, disease free survival, and larynx preservation.
If possible, outcome data were separately divided into each subgroup.According to the age were lower or higher than the median of the entire trials, it was classi ed as "low" or "high" subgroup.The prede ned primary endpoint was overall survival (OS) at longest follow-up available.The prede ned secondary outcomes were disease free survival (DFS) at longest follow-up available, and larynx preservation.

Quality assessment
Risk of bias of the eligible trials was independently assessed by two authors, with no a liation with any of the eligible trials.In cases of eventual divergences by discussion to achieve a consensus or consultation with a third author.Based on the Cochrane Collaboration methods 37 , the methodological quality of randomized controlled trials (RCTs) was assessed.The eligible studies were classi ed as low quality, or high quality according to the following criteria: 1) when allocation concealment or randomization was considered as a high risk of bias, regardless of the risk of other items, studies were assessed low quality; 2) if both allocation concealment and randomization were considered as a low risk of bias, and all other items were considered as low or unclear risk of bias in a study, studies were considered high quality.Based on the Newcastle-Ottawa scale, the study quality of observational studies was evaluated.The quality of trials was adjudicated on three broad perspectives: selection, comparability, and exposure/outcome, with four items (one star each), one item (up to two stars) as well as three items (one star each), respectively.In this analysis, studies were judged as high quality if they scored ≥ 7 stars.

Statistical analyses
For dichotomous outcomes, relative risk (RR) with 95% con dence interval (CI) was calculated.Between-study heterogeneity by the estimates of diversity (D 2 ), the inconsistency statistics (I 2 ) and visual inspection of the forest plots was evaluated.If a P value < 0.05 or I 2 > 50%, substantial heterogeneity was considered existing.We adopted a xed effects model and when I 2 = 0.If I 2 was above zero, we adopted both xed and random effects models, and used the most conservative estimate being the estimate with the widest CI or the point estimate closest to no effect.Pre-speci ed subgroups were performed based on study location (western country vs. eastern country), study design (randomized controlled trial vs. cohort), age (> 60 vs.≤ 60), and study quality (high quality vs. low quality).Analysis was conducted to evaluate whether the difference was statistically signi cant between the subgroups.In additional, subgroup differences were tested by Chi-square test-that is, whether the observed differences are aligned with chance alone within the subgroups.A low P value or a large Chi-square statistic relative to its degree of freedom construct evidence of heterogeneity beyond chance.
Meta-regression analysis was conducted to explore the associations between the primary outcome and characteristics of participants.Factors considered variables included publication year, percentage of male, mean age, simple size, percentage of T4 stage, percentage of positive lymph node, percentage of stageIV, percentage of pyriform sinus, and mean follow-up time.We evaluated publication bias for primary outcomes by funnel plot inspection and the Harbord test, in which 10 or more studies provided data.
All meta-analyses were conducted with RevMan version 5.3 (The Cochrane Collaboration, Copenhagen, Denmark) and Stata version 14 (StataCorp, College Station, Texas, USA).All P values are double-tailed, and the statistical signi cance was set at 0.05.

Grading the quality of evidence
Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach was used to assess the overall quality of evidence for each estimated outcome.According to study design, study quality, precision, consistency, directness and the risk of reporting bias, the overall quality of evidence is graded "high", "moderate", "low" or "very low" for each outcome.The analyses was performed in GRADE Pro-version 3.6 software.

Trial sequential analysis
We conducted trial sequential analysis (TSA) to explore whether cumulative data were adequately powered to assess outcomes.In this procedure, according to an alpha value set at 5% to determine signi cance, we established Z-curves for the primary outcome and secondary outcomes.Using the O'Brien-Fleming alpha spending method, we constructed adjusted signi cance trial sequential monitoring boundaries, with the hypothesis that a new study was successively added to the meta-analysis when signi cant testing may have been conducted each time.
Based on a pre-de ned relative risk reduction or increase of 15% in each outcomes, the required information size was computed, with the type I error (α) at 0.05 or 0.01 and the power (1 -β) at 0.90.TSA software version 0.9.5.9 (Centre for Clinical Intervention Research) was used for these analysis 38 .

Identi cation of studies
Figure 1 shows the owchart for the research selection program.Initially, the PubMed search identi ed 2,341 trials, the ISI Web of Knowledge 2,760, Cochrane Library 552, and Embase 442.After 3083 were removed as duplicates, 2924 trials were further excluded after titles and abstracts reviewed, and 88 literatures were scrutinized by scanning their full texts.Ultimately, 17 researches 11 − 27 met our prede ned criteria and were included in the meta-analyses.

Characteristics of the eligible studies
Table 1 present the main characteristics of selected studies.The 17 studies included two RCTs 11,22 and fteen retrospective cohorts 12 − 21,23−27 , with a total of 2539 patients, and were published from 1994 to 2019.Eight trials were conducted in western countries (eg.France, USA, Croatia, Portugal, The Netherlands), eight in eastern countries (eg.Taiwan, Korea, China), one not available.The simple size varied from 47 to 857 per study, the average age of patients varied between 54.4 and 64 years.Percentages of man, percentage of T4 stage, percentage of positive lymph node, percentage of stageIV and percentage of pyriform sinus, varied widely, with ranging from 78.6-100%, ranging from 4.3-86.7%,ranging from 63.4-91.8%,ranging from 6.6-87.8%,ranging from 50.6-100%, respectively.Mean follow-up periods varied, ranged from 19-92 months.

Methodological quality of eligible studies
Figures S1 and S2 shows the quality assessment of the RCTs.To summarize, neither two RCTs reported the blinding of patients and personnel, or the blinding of outcome assessment.On account of other items in RCTs was low bias, two RCTs were adjudicated high quality.Table S2 displays the quality assessment of the cohorts.These results showed that ve studies, which was rated as high quality, scored between 7 and 8 points, and the other ten studies, which was considered as low quality, scored below 6 points.

Primary outcomes
Results of the pooled effect indicated that the difference between non-surgical treatment and initial surgical on overall survival (OS) was not statistically of signi cance (relative risk [RR] 1.04; 95% con dence interval [CI] 0.94 to 1.15; I 2 = 5%) (Fig. 2A), and the summary OR and 95% CI were estimated using the random effect model with the widest CI.TSA results indicated that the cumulative Z-curve did not enter the futility area (Fig. 2B).The estimated required information size to cross the futility boundaries was 3823 randomized patients.The subgroup analysis did not nd any signi cant differences between subgroups on the basis of study location, study design, age, and study quality (

Secondary outcomes
Pooled analysis demonstrated patients treated with primary surgical experienced an increased disease free survival (DFS) probability compared with non-surgical treatment, with a RR of 1.20 (95% CI = 1.05 to1.37;I 2 = 0%) (Fig. 3A).However, the results of TSA indicated that the cumulative Z-curve just reached the conventional boundary, and did not cross the trial sequential monitoring boundary (Fig. 3B).The estimated required information size to cross the boundary for conventional and trial sequential monitoring was 3367 randomized patients.
Pooled estimates indicated compared to non-surgical treatment, initial surgical signi cantly lowered larynx preservation, with a RR of 0.48 (95% CI = 0.33 to 0.70; I 2 = 73%) (Fig. 4A).TSA analysis shown the cumulative Z-curve crossed the boundary for conventional and trial sequential monitoring (Fig. 4B).

Meta-regression analysis, publication bias, and GRADE
A meta-regression analysis shown that simple size signi cantly in ected effect size (P = 0.038) (Fig. 5D), which a trend towards a greater and more conform declines in survival probability was observed among patients received surgical management.Whereas year of publication (P = 0.402), percentage of male (P = 0.931), mean age (P = 0.893), percentage of T4 stage (P = 0.941), percentage of positive lymph node (P = 0.556), percentage of stageIV (P = 0.278), percentage of pyriform sinus (P = 0.144), and mean follow-up time (P = 0.556) did not signi cantly alter effect size, respectively (Fig. 5A, B, C, E, F, G, H and I).There was no evidence of publication bias for the primary outcome (OS) on inspection of the funnel plot (FigureS3) or from the Harbord test, 0.67.The GRADE level of evidence for OS and larynx preservation was very low and the level of evidence for DFS was low.The results was presented in the 'summary of ndings' (Table 3).*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes.The corresponding risk (and its 95% con dence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).CI: Con dence interval; RR: Risk ratio; GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our con dence in the estimate of effect.Moderate quality: Further research is likely to have an important impact on our con dence in the estimate of effect and may change the estimate.Low quality: Further research is very likely to have an important impact on our con dence in the estimate of effect and is likely to change the estimate.Very low quality: We are very uncertain about the estimate.

Discussion
The current meta-analysis, which involved two RCTs, comprising 286 patients, and fteen retrospective cohorts, including 1987 patients, compared primary surgical with non-surgical administration.The summary results con rmed that between initial surgical and non-surgical management have similar overall survival (RR 1.04, CI = 0.94 to1.15;I 2 = 5%), with TSA inconclusive.Subgroup analysis veri ed that the nding were generally conform, regardless of study location, study design, age, and study quality.Additionally, compared to non-surgical treatment, the primary surgical signi cantly improved disease free survival (RR 1.20, CI = 1.05 to1.37;I 2 = 0), with TSA inconclusive.Noteworthily, non-surgical administration did have a bene cial effect on larynx preservation, with TSA conclusive.
Recently systematic reviews illustrated management of advanced hypopharyngeal carcinoma 28 , while it did not statistically integrate available data to analyze whether there exist signi cant e cacy difference between primary surgical and non-surgical therapy or not.Meanwhile, it did not incorporate three additional new trials 12,19,17 , with two of them as lager simple retrospective cohorts.In 2018, Kim et al., using the latest population-based database (Surveillance, Epidemiology, and End Results (SEER)), analyzed the comparative effectiveness of primary surgical and non-surgical therapy for advanced hypopharyngeal cancer in the contemporary era, and indicated that treatment outcomes in the chemoradiotherapy group were comparable to that of the surgical group for locally advanced hypophagyngeal cancer 19 .Sequentially, Chung et al. performed a multicenter trials comparing induction chemotherapy with (chemo) radiotherapy, de nitive chemoradiotherapy and surgerybased therapy to assess the relative e cacy outcomes on patients with locally advanced hypopharyngeal squamous cell carcinoma, which veri ed that whether induction chemotherapy followed by (chemo) radiotherapy or de nitive chemoradiotherapy showed a comparable overall survival rate to surgery-based therapy 12 .
To the best of our knowledge, this is rst comprehensive meta-analysis to summarize all current studies focusing on comparing the relative effectiveness of between primary surgical and non-surgery therapy in advanced hypopharyngeal carcinoma.The meta-analysis and systematic review have a number of methodological strengths.Firstly, the protocol of this study was registered on PROSPERO.A prior-registered protocol may enhance the quality and transparency of meta-analyses.Secondly, prede ned subgroup was performed to explore that speci c patients can bene t from initial surgical or non-surgical treatment.Additionally, meta-regression was conducted to investigate the potential modi ers on outcomes.Thirdly, TSA was used to evaluate the risk of random errors (false discoveries), and its results support the contention that a 15% relative decrease or increase in incidence of events can be con dently removed.Finally, we analyzed the level of evidence using the GRADE method, which ranks evidence of clinical outcomes as high, moderate, low, or very low quality.
Our meta-analysis con rmed that the primary surgical compared to non-surgical management does not signi cantly improve overall survival in advanced hypopharyngeal cancer.However, our TSA presented that the Z-curve not only did not cross the conventional boundary and trial sequential monitoring boundary for bene t, but also did not enter the futility area, which show that conclusive evidence is not established and more researches were needed to a rm the results.Subgroup analysis veri ed that the result were generally consistent, regardless of study location, study design, age, and study quality.In additional, we found that the primary surgical can increase disease free survival probability compared with non-surgical administration, whereas TAS result showed that the cumulative Z-curve did not cross the trial sequential monitoring boundary.The result indicated that the required information size was not reached, the evidence achieving a credible conclusion was insu cient and future clinical researches were needed to a rm.It is worth mentioning that non-surgical management did have a positive effect on larynx preservation, with TSA conclusive.Regarding larynx preservation, this evidence is rm conclusion, and that further trials are not required.
A meta-regression analysis using restricted maximum likelihood estimators nd that simple size had potential in uence on treatments when the simple size in a trial was used as a moderating variable.Simple size possess a signi cant declining trend on effect size, that is to say, the primary surgical will be inferior to non-surgical treatment with the increase of sample size in advanced hypopharyngeal carcinoma.The nding may be explained that Kim et al. trial with maximum sample size in all included studies account for 22.1% overall estimate from forest plot and exert an observable impact on effect size.What's more, the result imply that further more trials are required, in accordance with the results of our TSA with regard to overall survival and disease free survival.Hence, further large and perspective RCTs were need to clarify the relative e cacy between surgery-based therapy and non-surgical treatment on survival outcome.
Currently, in the context of the global trend toward organ-preserving treatment (non-surgical treatment), de nitive chemoradiotherapy has obtained increasing prevalence as the de nitive therapy in hypopharyngeal cancer.Given we cannot come to a clear conclusion that non-surgical treatment is non-inferior to surgery-based therapy in terms of survival outcome, thus it is of great importance to identify reliable and convenient clinical factors/markers to screen out which subset of patients with advanced hypopharyngeal cancer is suitable to organ-preserving treatment using de nitive chemoradiotherapy with surgery reserved for salvage, and optimize non-surgical therapeutic strategies.Induction chemotherapy appears to serve as mediator, as studies revealed that a response to incipient chemotherapy predicted a response to subsequent treatment in the advanced head and neck cancer, including oral cavity, oropharynx, laryngeal and hypopharynx 29 − 31 .In additional, compared to non-responders, responders had signi cantly favorable survival.The concept was also adopted in the Veterans Affairs Laryngeal Cancer Study 7 .On account of hypopharynx cancer accounting for minor part in above trials, the effect of response to induction chemotherapy only in advanced hypopharynx cancer patients to determine organ-preserving treatment are awaited to further research.In recent years, human papilloma virus (HPV) infection have been recognized as risk factors for HNC, including hypopharynx.Several trials indicated that HPV positive patients were sensitive to radiotherapy and chemotherapy as well as showed superior survival 32,33 .Maybe HPV status can be a predictor to select non-surgical treatment.
Future larger prospective studies with adequately powered are warranted to clarify issues.
Several limitations should be noted in our study.First, the number of retrospective cohorts was relatively large, and there were only two old RCTs published in the early 1980s.As for retrospective trials, selection bias is very likely due to no randomisation between the treatment strategies.In regard to RCTs, neither trials stated blinding for participants, personnel or the outcome assessment, which augmented the bias risk of performance and detection.On the basis of above reasons, GRADE judged the certainty of the body of evidence as low or very for each outcome.
Hence, there is a need for large, multi-centered RCTs to support the current results.Second, though we found that non-surgical administration can remarkably improve larynx preservation, the obvious heterogeneity was observed.This is because marked heterogeneity existed across the included studies in terms of the different types of chemotherapy received, as well as the type and doses of radiotherapy.But almost all studies there is a clear superiority with non-surgical treatments in terms of preserving a functional larynx.Third, a large proportion of eligible trials had a small sample size, with only 6 of the 17 trials containing a sample size of larger than 100 patients.Just as our ndings of meta-regression analysis and TSA, the small sample sizes may have in uence on treatment effects, indicating a need for large and perspective researches.
Fourth, the safety of non-surgical administration cannot be detected on account of insu cient data.Most of the included studies did not thoroughly evaluated the adverse effect of organ preservation treatment in locally advanced hypopharyngeal cancer, unlike in carcinoma of the larynx, which hindered a de nite conclusion regarding optimal treatment regimen 34 .Further well-designed researches are warranted to resolve these issues.

Conclusion
The results of our meta-analysis indicated treatment outcomes in non-surgical therapy were comparable to that of surgery-based therapy for advanced-stage hypopharyngeal cancer, which makes larynx preservation feasible.It can be served as an alternative approach in patients who are un t for or refuse surgery.On account of inconclusive evidence from TSA, further prospective investigations comparing modern non-surgical management with modern surgical techniques remain necessary.FigureS3.Assessment of publication bias using a funnel plot.

Abbreviations
TableS1.Treatment modalities of the studies included in the meta-analysis.
TableS2.Quality Assessment With Newcastle Ottawa Scale for cohort study TableS3.Meta-regression analysis analyzing the potential confounders for outcome

Supplementary Files
This is a list of supplementary les associated with this preprint.Click to download. PRISMAchecklist.doc

Table S1
details treatment modalities of the studies included in the meta-analysis.

Table 1
Baseline characteristics of the studies included in the meta-analysis.

Table 3 Summary
TSA= trial sequential analysis, GRADE= The Grading of Recommendations Assessment Development and Evaluation, RR= relative risk, CI=con dence interval, OS=overall survival, DFS= disease free survival, WHO=World Health Organization, HNC=head and neck cancers, PRISMA=Preferred Reporting Items for Systematic Reviews and Meta-analyses, PICOS =participants, interventions, comparators, outcomes, and study design, RCTs= randomized controlled trials, SEER=Surveillance, Epidemiology, and End Results, HPV= human papilloma virus, NA=not available, SD=standard deviation, No.= number of participants, RC=retrospective cohort, RT= radiotherapy, CRT= chemoradiotherapy, CT= chemotherapy, REC=representative of exposed cohort , SNC=selection of nonexposed cohort, AE=ascertainment of exposed, AOI=absence of outcome of interest.