Adjuvant Systemic Therapy after Chemoradiation and Brachytherapy for Locally Advanced Cervical Cancer: A Systematic Review and Meta-Analysis

Simple Summary The standard of care for locally advanced cervical cancer is chemoradiation and brachytherapy. The addition of adjuvant systemic treatment may improve overall survival. A systematic review and meta-analysis were conducted to summarize evidence on survival outcomes, treatment completion and toxicity. Thirty-five articles reporting on 29 different studies were selected from a total of 612 articles published on this topic since 2000. Twelve studies on two different chemotherapy combinations (platinum–pyrimidine antagonist and platinum–taxane) were included for meta-analysis. Both these adjuvant chemotherapy combinations did not yield a survival benefit but did lead to more severe side-effects than chemoradiation only. Therefore, these adjuvant treatment strategies cannot be recommended for unselected patients with locally advanced cervical cancer. Most of the studies on other chemotherapeutic agents did not seem to provide a good balance between efficacy and toxicity either. The evidence on adjuvant immunotherapy for locally advanced cervical cancer is still immature. Abstract Background: Standard of care for locally advanced cervical cancer is chemoradiation and brachytherapy. The addition of adjuvant systemic treatment may improve overall survival. A systematic review and meta-analysis was conducted to summarize evidence on survival outcomes, treatment completion and toxicity. Methods: PubMed, EMBASE and Web of Science were systematically searched for relevant prospective and retrospective studies. Two authors independently selected studies, extracted data and assessed study quality. Pooled hazard ratios for survival endpoints were estimated using random effect models. Weighted averages of treatment completion and toxicity rates were calculated and compared by the Fisher exact test. Results: The search returned 612 articles; 35 articles reporting on 29 different studies on adjuvant chemotherapy or immunotherapy were selected for systematic review. Twelve studies on an adjuvant platinum–pyrimidine antagonist or platinum–taxane were included for meta-analysis. The pooled hazard ratios for overall survival were 0.76 (99%CI: 0.43–1.34, p = 0.22) and 0.47 (99%CI: 0.12–1.86, p = 0.16) for the addition of, respectively, a platinum–pyrimidine antagonist or platinum–taxane to chemoradiation and brachytherapy. Completion rates were 82% (95%CI: 76–87%) for platinum–pyrimidine antagonist and 74% (95%CI: 63–85%) for platinum–taxane. Severe acute hematological and gastro-intestinal toxicities were significantly increased by adding adjuvant chemotherapy to chemoradiation and brachytherapy. Conclusions: The addition of adjuvant platinum–pyrimidine antagonist or platinum–taxane after chemoradiation and brachytherapy does not significantly improve overall survival, while acute toxicity is significantly increased. These adjuvant treatment strategies can therefore not be recommended for unselected patients with locally advanced cervical cancer.


Introduction
Cervical cancer is the second most common cancer in women across the world [1]. The standard of care for locally advanced cervical cancer has been platinum-based chemoradiation with brachytherapy since the National Cancer Institute alert in 1999 [2]. A metaanalysis of 13 randomized controlled trials showed a 6% improvement in 5-year overall survival by adding concurrent chemotherapy to radiation [3]. In recent years, further improvement of overall survival was reported with image-based brachytherapy and radiation dose escalation, while reducing toxicity [4][5][6][7][8]. These treatment advances changed patterns of failure. Distant metastases are now the most common type of failure, occurring in 24-30% at 5 years after chemoradiation and brachytherapy [9,10]. Distant metastases occur due to the incomplete eradication of the primary tumor or involved lymph nodes or due to undetected micro-metastasis outside the field of treatment [9]. Adjuvant systemic therapy after chemoradiation and brachytherapy has the potential to reduce the risk of distant metastasis and improve overall survival.
A sub-analysis of the aforementioned meta-analysis showed that concurrent chemoradiation with adjuvant chemotherapy yielded a 19% 5-year overall survival benefit compared to radiotherapy alone [3]. Since chemoradiation is nowadays the standard of care, an important question is whether and to what extent overall survival is improved by chemoradiation followed by adjuvant systemic therapy compared to chemoradiation. A randomized controlled trial by Duenas-Gonzales et al. on radiotherapy with concurrent cisplatin vs. radiotherapy with concurrent cisplatin-gemcitabine followed by adjuvant cisplatin-gemcitabine showed a significant improvement in overall survival (hazard ratio (HR) 0.68, 95% confidence interval (CI): 0.49-0.95, p = 0.02). However, this trial could not give an answer as the treatment groups did not receive the same concurrent chemotherapy [11]. A 2014 Cochrane review was also not able to answer this question because only two randomized controlled trials on chemoradiation followed by adjuvant chemotherapy vs. chemoradiation were found and not pooled [12].
An overview of all clinical studies on adjuvant systemic therapy after chemoradiation and brachytherapy is therefore needed to summarize the impact, if any, on disease-related outcomes and to provide direction for the design of future trials. We performed a systematic review and meta-analysis to provide this overview and pooled estimates of the efficacy and toxicity of adjuvant systemic therapy after chemoradiation and brachytherapy for locally advanced cervical cancer.

Article Characteristics
The design of the systematic search is presented in Table 1. Articles on randomized and non-randomized prospective and retrospective studies were eligible if chemoradiation with brachytherapy followed by adjuvant systemic therapy was investigated or compared to standard chemoradiation with brachytherapy. The following article types were not eligible: conference abstracts, case-reports, review articles, meta-analyses, editorials, letters to editor and guidelines.

Patient
Tumor characteristics: FIGO stage IB-IVA (including metastasis to the para-aortic lymph nodes) cervical cancer of squamous cell carcinoma, adenocarcinoma or adenosquamous carcinoma histotype Study characteristics: randomized controlled trials, non-randomized prospective and retrospective studies

Intervention
External beam radiotherapy to the whole pelvis (with or without integrated or sequential boosts or extended field) with concurrent chemotherapy and intracavitary or interstitial brachytherapy followed by adjuvant systemic therapy (e.g., chemotherapy or immuno therapy)

Control
External beam radiotherapy to the whole pelvis (with or without integrated or sequential boosts or extended field) with concurrent chemotherapy and intracavitary or interstitial brachytherapy

Outcomes
Overall survival, recurrence-or disease-free survival, metastasis-free survival, treatment completion, toxicity Articles published before the year 2000 were excluded because concurrent chemoradiation was not established as the standard of treatment before the National Cancer Institute alert [2].

Study Population Characteristics
Patients with a diagnosis of Fédération Internationale de Gynécologie et d'Obstétrique (FIGO) stage IB-IVA cervical cancer of the squamous cell carcinoma, adenocarcinoma or adenosquamous carcinoma histological type were included.
Studies including patients with proven or suspected metastasis to para-aortic lymph nodes were eligible. Studies on patients who were treated with primary surgery and adjuvant chemoradiation and systemic chemotherapy were not included. Neither did we include studies of patients with distant metastasis or with persistent or recurrent cervical cancer after failure of previous treatment(s).

Treatment Characteristics
Studies were eligible if external beam radiotherapy was delivered to the whole pelvis with or without integrated or sequential boost(s). Extended field external beam radiotherapy for involved or suspected para-aortic lymph nodes or as prophylactic treatment was allowed. Concurrent chemotherapy was preferably platinum-based, but other concurrent agents were accepted. During or after chemoradiation, patients had to undergo intracavitary and/or interstitial brachytherapy. Adjuvant systemic therapy had to consist of at least one administration after chemoradiotherapy and brachytherapy of any systemically active agent, e.g., chemotherapy, monoclonal antibodies, immunomodulators, and the intent of treatment had to be curative.

Outcomes Measures
Studies had to report on ≥1 of the following outcomes: distant metastasis-free survival; recurrence or disease-free survival; overall survival; treatment completion; and toxicity. These outcomes had to be reported separately for the patients undergoing chemoradiation and those undergoing chemoradiation followed by adjuvant systemic therapy as applicable.

Literature Searches
Search strings for PubMed, Web of Science and EMBASE were devised by N.H. with assistance from a trained librarian (provided in Supplementary Materials I), to identify relevant studies published until 5 September 2020. The following terms, and possible variations thereof, were matched to appropriate medical subject headings: "cervical cancer"; "chemoradiation" and "adjuvant therapy". Searches were restricted to publications in or after the year 2000. Study authors were contacted if full texts were not available. Grey literature sources, such as clinicaltrials.gov and Google Scholar, were searched for ongoing and unpublished trials.

Selection of Studies
All articles were imported in EndNote X9 and deduplicated before study selection. Two reviewers (N.H. and P.M.) independently read the titles and abstracts of all articles to identify relevant studies for full text review. Hand searches of reference lists of the articles selected for full text review were performed to identify additional relevant articles. At every stage of the selection process, the independent results of the two review authors were compared and any differences were solved in consensus meetings or by the decision of a third reviewer (S.C.). All selected studies were included in the systematic review. The inclusion in the meta-analysis was possible if there were two or more articles on chemoradiation and brachytherapy vs. chemoradiation and brachytherapy followed by a systemic agent (combination).
If ≥1 article described the same study, the most recent and complete article was used for analysis. However, if efficacy and toxicity outcomes were reported in two separate articles, both were included.

Risk of Bias Assessment
A pre-specified risk of bias assessment (provided in the Supplementary Materials) was based on the Cochrane Handbook for Systematic Reviews version 5.1.0 [13], and the "Meta-analysis Of Observational Studies in Epidemiology" consensus on reporting of observational studies [14]. Both review authors (N.H. and P.M.) were trained and a pilot study with a test article was conducted. For each study, the two reviewers independently rated all risk of bias aspects. Studies with discrepancies were listed and discussed, and in case of remaining disagreement, the third reviewer was consulted. A study's overall risk of bias was classified as: (1) low if the risk of bias was low for all domains; (2) some concerns if there were unclarities or some concerns of risk of bias in one domain; (3) high if there was a high risk of bias in ≥1 domain. The overall risk of bias will be reported along with the outcomes of the included studies.

Data Extraction
The data extraction protocol is provided in the Supplementary Materials I. The following pre-specified information was extracted: publication details, study design and population, treatment and summary measures of outcomes. The latter consisted of followup time, treatment completion rates, survival outcomes (2-and 3-year estimates, hazard ratios with 95% confidence intervals, acute (occurring <3 months) and late (occurring or persisting beyond 3 months) severe toxicity (grade 3-5). If survival outcomes were not directly reported, the estimates were deducted from survival graphs or reported crude numbers. Extracted data were compared, variables with discrepancies were listed and discussed by the two reviewers and with the third reviewer in case of disagreement.

Statistical Methods
The primary outcome is overall survival. Secondary outcomes are distant metastasisfree survival and recurrence-free survival, treatment completion rate and the rates of severe acute and late toxicities. Overall survival time is defined as the time from the date of inclusion/randomization to date of death, or date of last follow-up in alive patients. Recurrence-free survival time is defined as the time from the date of inclusion/randomization to the date of first recurrence (regardless of localization), or the date of last follow-up in patients without recurrence. Metastasis-free survival time is defined as the time from date of inclusion/randomization to date of first distant metastasis (recurrence beyond pelvis or para-aortic lymph node chain), or the date of last follow-up in patients without recurrence. Severe toxicities were defined as grade ≥3 according to the Common Terminology Criteria for Adverse Events or the Radiation Therapy Oncology Group/European Organisation for Research and Treatment of Cancer classification, as reported in the original articles.
Total radiotherapy doses were calculated as doses equivalent to 2 Gy fractions (EQD2) using an α/β = 10. Pooled estimates of radiotherapy dose, treatment completion and toxicity rates were calculated as weighted averages with 95% CIs and compared with the Fisher exact test.
For the survival endpoints, if included studies reported HRs and 95% CIs, the natural logarithm of HR and its variance were calculated directly (Supplementary Materials I). If not, these were imputed using other data provided in the article according to the methodology of Tierney et al. (Supplementary Materials I) [15]. Pooled estimates of the hazard ratios for overall survival, distant metastasis-free survival and recurrence-free survival were calculated using random effects models (DerSimonian-Laird method) wherein each study is weighed according to their sample size. Separate models were built for each adjuvant systemic treatment for which ≥2 studies reported survival outcomes. Randomeffects models were chosen a priori because of the anticipated clinical and methodological heterogeneity between the studies. The level of statistical significance was pre-defined as p < 0.01 to correct for multiple testing. Heterogeneity, in effect size among studies, was assessed by the I 2 and the Q-test. Significant statistical heterogeneity between studies was defined as an I 2 > 50% with the Q-test p < 0.05. Heterogeneity due to of pooling studies with different designs was addressed by pre-specified subgroup analyses (randomized controlled trials vs. non-randomized controlled trials). Pre-specified sensitivity analyses consisted of re-estimating all pooled estimates according to the leave-one-out method using random effect models, to evaluate whether the results could have been affected markedly by a single study. Publication bias was assessed using funnel plots in the meta-analysis for the primary endpoint. Descriptive analyses were used if the data were limited.
Analyses were performed in Microsoft Excel and R version 3.6.1 (http://www.rproject.org/ (accessed on 8 April 2021)). R packages used in this study are reported in the Supplementary Materials.
This study was registered at PROSPERO under registration number CRD42020211194 and conducted according to the Preferred Reporting Items for Systematic Review and Meta-Analysis guidelines [16].

Systematic Searches
Systematic searches yielded 612 unique articles ( Figure 1). Forty-nine were selected for full-text review, of which 32 were eligible for inclusion. Hand searches of reference lists yielded another three eligible articles. These 35 articles reported on 29 different studies and were included in the systematic review. Twelve of 29 reported studies were also included in the meta-analysis. Reasons for the exclusion of the remaining 17 studies are listed in Figure 1. Table 2 shows the main characteristics of the 29 studies included in the systematic review. Fifteen of the 29 included studies compared chemoradiation followed by adjuvant systemic therapy to chemoradiation alone using various designs. The remaining 14 studies  Table 2 shows the main characteristics of the 29 studies includ review. Fifteen of the 29 included studies compared chemoradiation f systemic therapy to chemoradiation alone using various designs. Th ies that reported on chemoradiation followed by adjuvant systemic t a control group treated with chemoradiation.

Characteristics Included Studies
Tables S1 and S2 in supplemental data II describe the radiother in the included studies. Generally, external beam radiotherapy planned using computed tomography and delivered by parallel op  Tables S1 and S2 in supplemental data II describe the radiotherapy techniques used in the included studies. Generally, external beam radiotherapy was conventionally planned using computed tomography and delivered by parallel opposing or box techniques. The use of extended field external beam radiotherapy for positive para-aortic lymph nodes was reported in 45% of the studies and prophylactic extended field was reported in 17% of the studies. Brachytherapy was most frequently radiograph-based using standard plans prescribing to point A and delivered with intracavitary applicators using high dose-rates. The use of interstitial needles and magnetic resonance imaging/computed tomographybased planning were quite uncommon. The cumulative (external beam radiotherapy + brachytherapy) prescribed EQD2 dose range was 78 Gy (95%CI: 76-81 Gy) to 88 Gy (95%CI: 87-90 Gy). Radiotherapy was completed in 91% (95%CI: 90-93%) in patients treated with chemoradiation followed by adjuvant systemic therapy and in 94% (95%CI: 93-95%) of patients treated with chemoradiation (p = 0.006).
Tables S3-S5 in supplemental data II shows the agents, doses and schedules of concurrent and adjuvant therapies. Concurrent chemotherapy was mainly platinum-based (93%). In 10 studies, a second agent was added, usually a pyrimidine antagonist. Five of the 12 (42%) controlled studies did not use the same agent(s) as concurrent treatment in the control and experimental arm [11,19,23,24,27].

Risk of Bias Assessment
Results of the risk of bias assessment are presented in Figure 2. All studies included in the meta-analyses on adjuvant platinum-pyrimidine antagonist were at high risk of bias. Systematic differences between the study arms leading to the performance bias were present in all six studies. In addition, the retrospective studies were at risk of selection bias, reporting bias, registration bias and confounding by indication. Likewise, five retrospective studies on adjuvant platinum-taxane were classified as at a high risk of bias. Only the randomized controlled trial on adjuvant platinum-taxane by Tangjitgamol et al. was judged to be at low risk of bias [30].
Meta-analyses were also performed for six controlled studies (total N = 622) reporting survival outcomes on adjuvant platinum-taxane ( Figure 3). The pooled hazard ratio for overall survival was 0.47 (99%CI: 0.12-1.86, p = 0.16). Heterogeneity (Supplemental Materials II Table S6) and publication bias (Figure 4) are present. The pooled hazard ratio for recurrence-free survival was 0.68 (99%CI: 0.33-1.41, p = 0.17). No meta-analysis could be performed for distant metastasis-free survival as only one study reported on this outcome (HR 0.26, 99%CI: 0.05-1.49, p = 0.047) (35). bias. Systematic differences between the study arms leading to the performance bias were present in all six studies. In addition, the retrospective studies were at risk of selection bias, reporting bias, registration bias and confounding by indication. Likewise, five retrospective studies on adjuvant platinum-taxane were classified as at a high risk of bias. Only the randomized controlled trial on adjuvant platinum-taxane by Tangjitgamol et al. was judged to be at low risk of bias [30].

Study
Year of publication  survival outcomes; discrepancies between reported estimates and survival curves.

Meta-Analysis
Meta-analyses were performed for six studies comparing chemoradiation followed by adjuvant platinum-pyrimidine antagonist with chemoradiation ( Figure 3). The pooled hazard ratio estimate for overall survival was 0.76 (99%CI: 0.43-1.34, p = 0.22). Heterogeneity (Supplemental Materials II Table S6) and publication bias (Figure 4) are present. Figure 3. Impact on the overall survival of the addition of adjuvant chemotherapy to chemoradiation and brachytherapy. Each study in the forest plot is represented by a black square which represents the study's hazard ratio and a whisker on each side that represents the study's 99% confidence interval. The size of the black square represents the weight of the study in the meta-analysis. The pooled hazard ratios are shown as diamond shapes; the light blue diamond represents the pooled hazard ratio based on only randomized controlled trials, the dark blue diamond represents the pooled hazard ratio based on only non-randomized studies, and the black diamond is the pooled hazard ratio of all studies combined. (A) Meta-analysis of overall survival after concurrent chemoradiation and brachytherapy with adjuvant platinum derivate and pyrimidine antagonist vs. concurrent chemoradiation and brachytherapy only. I 2 = 62%, Q-test p = 0.02. (B) Meta-analysis of overall survival after concurrent chemoradiation and brachytherapy with adjuvant platinum Each study in the forest plot is represented by a black square which represents the study's hazard ratio and a whisker on each side that represents the study's 99% confidence interval. The size of the black square represents the weight of the study in the meta-analysis. The pooled hazard ratios are shown as diamond shapes; the light blue diamond represents the pooled hazard ratio based on only randomized controlled trials, the dark blue diamond represents the pooled hazard ratio based on only non-randomized studies, and the black diamond is the pooled hazard ratio of all studies combined.   Sensitivity analysis based on the leave-one-out approach showed that outcomes of the meta-analyses for the primary outcome are robust (Supplemental Materials II Table S7).

Systematic Review of Survival Outcomes
An overview of the survival outcomes reported in all 29 included studies are provided in Tables S8 and S9 of the Supplementary Materials II. Briefly, studies on adjuvant platinum-pyrimidine antagonist after chemoradiation reported 3-year overall survival rates of 70-90% in the experimental arm compared to 69-93% in the control arm. Studies on adjuvant platinum-taxane showed 3-year overall survival rates of 31-80% in the experimental arm compared to 23-93% in the control arm. A randomized controlled trial on adjuvant 5-fluorouracil showed no significant benefit for metastasis-free, recurrence-free and overall survival [26]. A phase I trial on ipilimumab reported a 1-year progression-free survival and overall survival of 81% and 90%, respectively [41].

Systematic Review of Feasibility and Toxicity
The feasibility and severe toxicity of adjuvant systemic therapy are presented Tables 3-5. Pooled treatment completion rate was 79% (95%CI: 76-82%) for chemoradiation followed by adjuvant platinum-pyrimidine antagonist and 70% (95%CI: 64-86%) for chemoradiation followed by adjuvant platinum-taxane. Both adjuvant chemotherapy doublets caused more severe acute hematological and gastro-intestinal toxicities than chemoradiation alone. Adjuvant ipilimumab was completed in 86% and immune-mediated toxicity was observed in some patients [41]. Adjuvant pembrolizumab was completed in 100% despite severe gastro-intestinal toxicities and hypothyroidism in 13 and 4%, respectively [39].    [30] and the burden of increased toxicity, to change clinical practice. If the outcome of the OUTBACK trial is negative, then all published studies will agree that adjuvant platinum-taxane is not a good strategy to improve the survival of patients with locally advanced cervical cancer.

Discussion
This systematic review and meta-analysis evaluated the potential benefit and toxicity of adjuvant systemic therapy after primary chemoradiation and brachytherapy for locally advanced cervical cancer. Twenty-nine studies reporting on adjuvant chemotherapy and immunotherapy were included. The meta-analysis of 12 studies on two chemotherapy doublets (platinum-taxane and platinum-pyrimidine antagonists) showed no significant overall survival benefit while severe acute toxicity was significantly increased.
In this meta-analysis, two randomized controlled trials on the benefit of the addition of a platinum-pyrimidine antagonist after chemoradiation have been pooled. A significant benefit for distant metastasis-free survival was found. The benefit for overall survival (HR 0.73 95%CI: 0.50-1.06) did not reach statistical significance (p = 0.029, predefined as p < 0.01 in this meta-analysis). A lack of power could be the reason that this result did not reach significance. The addition of four non-randomized studies, which increased the numbers of patients from 670 to 1394, and did not result in a statistically significant overall survival benefit either (HR 0.76, 95%CI: 0.43-1.34, p = 0.22). These pooled estimates should be interpreted with caution, because several forms of bias may have affected trial outcomes. In both randomized controlled trials, the superior outcomes in the experimental arm may have partly been due to the addition of a pyrimidine antagonist to concurrent chemotherapy. In the non-randomized studies, the risk of recurrence and death may not have been in same between study arms. In addition, publication bias is probably present which may have biased the pooled overall survival estimate in favor of adjuvant platinumpyrimidine antagonist. Clearly, there is a need for a high quality randomized controlled trial that is powered to demonstrate the significance of a benefit around HR 0.75. It is not likely that such a trial will be conducted because all studies showed a substantial increase in severe toxicity. Therefore, adjuvant platinum-pyrimidine antagonist cannot be recommended for unselected patients.
Subgroup-analysis of the randomized controlled trial by Duenas-Gonzales et al. showed that patients with Stage III-IVA, tumors ≥5 cm and of non-adenocarcinoma histotype had benefitted the most from adjuvant treatment [20]. The contrary was found in the study by Fabri et al.: stage ≥IIIA had a significantly worse overall survival despite adjuvant platinum-pyrimidine in multivariable analysis [21]. Kim et al. found that adjuvant platinum-pyrimidine and tumor characteristics were not significantly related to overall survival and disease-free survival in univariable analyses [22]. The other studies on adjuvant platinum-pyrimidine did not report subgroup analysis, therefore no pooled estimates could be calculated for patients with additional risk factors. Hence, current evidence is unclear about a possible benefit in high-risk subgroups, but it is clear about the significant increase in severe toxicity.
The second meta-analysis pooled one randomized controlled trial (N = 259) with five small retrospective cohorts (N = 363) on adjuvant platinum-taxane. The randomized controlled trial was at low risk of bias and showed no benefit of adjuvant platinum-taxane for overall survival and recurrence-free survival. The addition of the five small studies (all at high risk of bias) did not change this conclusion. However, the pooled estimate would change if a new, large study was added. The outcome of the ongoing OUTBACK study (ANZGOG-0902/GOG-0274/RTOG-1174), a large phase III randomized controlled trial on adjuvant carboplatin-paclitaxel, is required for a conclusion on the benefit of adjuvant platinum-taxane [51]. As such, this meta-analysis could serve as an overview of the literature that facilitates the interpretation of the OUTBACK results.
Considering the toxicity profile, targeted therapies might be an attractive alternative to chemotherapy. Only two phase I-II studies on adjuvant immunotherapy have been published so far. The first by Mayadev et al. (2019) was a phase I trial on the anti-cytotoxic Tlymphocyte antigen-4 (CTLA-4) ipilimumab as an adjuvant agent after chemoradiation [41]. This treatment seems feasible (86% completion) but severe immune-mediated toxicities are not uncommon. The reported 1-year progression-free survival of 81% is difficult to interpret as it was not compared to chemoradiation. In metastatic cervical cancer, ipilimumab showed no significant clinical activity as monotherapy [52].
The second study was published in 2020 by Duska et al.: a phase-II randomized controlled trial on concurrent vs. adjuvant pembrolizumab (PD-1 inhibitor). In both settings pembrolizumab was feasible (83% completion) and severe toxicity was limited [39]. No data on efficacy was reported. The efficacy of pembrolizumab has been demonstrated for PD-L1-positive recurrent and metastatic cervical cancer which led to registration for this indication [53]. A phase-III randomized controlled trial (KEYNOTE-A18) on pembrolizumab during and after chemoradiation compared to chemoradiation is ongoing [54].
In addition to the addition of systemic therapies to chemoradiation, the improvement of radiotherapy techniques is also a way to increase tumor control while reducing toxicity. Image-guided adaptive brachytherapy has already been shown as able to do this [4][5][6][7][8], and new evidence on the impact of intensity-modulated radiotherapy and applying strict dose aims and constraints for external beam radiation therapy are expected to be published soon by, respectively, the investigators of the PARCER trial [55,56] and the EMBRACE II study [57].
This study was limited to adjuvant systemic therapy after chemoradiation and brachytherapy because this may be the most feasible option in terms of toxicity. It also allows for patient selection based on response to chemoradiation, which is an independent predictor of overall survival [58].
A couple of weaknesses should be considered when interpreting this meta-analysis. Firstly, there is inter-study heterogeneity which could affect the accuracy of the pooled estimates. For example, in a part of the included studies (prophylactic), extended field radiotherapy has been applied, which could have increased the risk of gastro-intestinal toxicity [59,60]. This is not surprising considering the variety of eligible study designs. We accounted for some heterogeneity by using random-effects models and we verified the robustness of the conclusions by sensitivity analysis. Secondly, the quality of the published studies provides suboptimal evidence. These limitations do not affect, but rather support the conclusion that adjuvant systemic therapy does not improve overall survival in unselected patients. Only if the ongoing OUTBACK study will show a substantial significant overall survival benefit, the conclusion of this meta-analysis could change [51].

Conclusions
The current standard of care for patients with locally advanced cervical cancer is cisplatin-based chemoradiation and brachytherapy. Improvement of overall survival of these patients has plateaued in the last two decades. This systematic review on the efficacy and toxicity of adjuvant systemic therapy gives an overview of current evidence. Few randomized controlled trials have been published and most studies were at considerable risk of bias. A meta-analysis of the two most investigated chemotherapy doublets showed that there was no significant improvement in overall survival while acute toxicity was significantly increased. Most of the studies on other chemotherapeutic agents did not seem to provide a good balance between efficacy and toxicity either. Current evidence on targeted therapies in the adjuvant setting is immature. Future clinical trials should be selective in the allocation of treatment strategy and focus on agents that increase the therapeutic window between efficacy and toxicity.  Table S6: Meta-analysis of survival outcomes and post hoc analysis; Table S7: Sensitivity analysis of survival outcomes; Table S8. Survival outcomes (1/2); Table S9. Survival outcomes (2/2).