FOLFIRINOX or Gemcitabine-based Chemotherapy for Borderline Resectable and Locally Advanced Pancreatic Cancer: A Multi-institutional, Patient-Level, Meta-analysis and Systematic Review

Background Pancreatic cancer often presents as locally advanced (LAPC) or borderline resectable (BRPC). Neoadjuvant systemic therapy is recommended as initial treatment. It is currently unclear what chemotherapy should be preferred for patients with BRPC or LAPC. Methods We performed a systematic review and multi-institutional meta-analysis of patient-level data regarding the use of initial systemic therapy for BRPC and LAPC. Outcomes were reported separately for tumor entity and by chemotherapy regimen including FOLFIRINOX (FIO) or gemcitabine-based. Results A total of 23 studies comprising 2930 patients were analyzed for overall survival (OS) calculated from the beginning of systemic treatment. OS for patients with BRPC was 22.0 months with FIO, 16.9 months with gemcitabine/nab-paclitaxel (Gem/nab), 21.6 months with gemcitabine/cisplatin or oxaliplatin or docetaxel or capecitabine (GemX), and 10 months with gemcitabine monotherapy (Gem-mono) (p < 0.0001). In patients with LAPC, OS also was higher with FIO (17.1 months) compared with Gem/nab (12.5 months), GemX (12.3 months), and Gem-mono (9.4 months; p < 0.0001). This difference was driven by the patients who did not undergo surgery, where FIO was superior to other regimens. The resection rates for patients with BRPC were 0.55 for gemcitabine-based chemotherapy and 0.53 with FIO. In patients with LAPC, resection rates were 0.19 with Gemcitabine and 0.28 with FIO. In resected patients, OS for patients with BRPC was 32.9 months with FIO and not different compared to Gem/nab, (28.6 months, p = 0.285), GemX (38.8 months, p = 0.1), or Gem-mono (23.1 months, p = 0.083). A similar trend was observed in resected patients converted from LAPC. Conclusions In patients with BRPC or LAPC, primary treatment with FOLFIRINOX compared with Gemcitabine-based chemotherapy appears to provide a survival benefit for patients that are ultimately unresectable. For patients that undergo surgical resection, outcomes are similar between GEM+ and FOLFIRINOX when delivered in the neoadjuvant setting. Supplementary Information The online version contains supplementary material available at 10.1245/s10434-023-13353-2.

Pancreatic ductal adenocarcinoma remains among the leading causes of cancer-related mortality. [1][2][3] Most patients present with an advanced stage of disease, e.g., distant metastasis, or locally advanced (LAPC) or borderline resectable (BRPC) pancreatic cancer, ultimately resulting in a dismal 10-15% upfront resection rate and poor 5-year survival. 2 In patients with BRPC or LAPC, primary chemotherapy with or without chemoradiation, hypofractionated radiotherapy, or stereotactic body radiotherapy has become increasingly utilized for local downstaging to enable a potentially curable resection. To this end, FOL-FIRINOX (FIO) and Gemcitabine-based regimens (GEM?) are commonly used for primary treatment of pancreatic cancer. [4][5][6] Generally, neoadjuvant FIO has been favored based on extrapolation of favorable outcomes in the adjuvant setting; however, this regimen carries a significant toxicity burden limiting use to patients with an excellent performance status. 7,8 Gemcitabine continues to be used for its moderate side effect profile and is commonly paired with other therapies (GEM?) with increased biological activity. [6][7][8][9] While PRODIGE 24 favored FIO over gemcitabine monotherapy in the adjuvant setting, these findings have yet to be demonstrated in the neoadjuvant setting with high level data. [6][7][8][9] Prospective series investigating this question are available, but often include a limited number of patients, and are usually highly heterogeneous due to the combination of patients with resectable tumors, BRPC, and LAPC, which limit the ability to assess the impact of the type of neoadjuvant chemotherapy on surgical resection rate, positive margin rate, and survival. 6,8,10,11 Thus, the objective of this study was to summarize the current evidence regarding initial systemic treatment for LAPC and BRPC through a multi-institutional meta-analysis of patient-level data from high-volume pancreatic centers. The primary goal of our meta-analysis was to assess the impact of primary chemotherapy, FIO, or GEM? regimens on survival rates of patients with BRPC and LAPC. Secondary outcomes included resection rates, R0 resection rates, and impact of radiotherapy.

Study Search and Selection
The systematic review and meta-analysis were conducted following the PRISMA guidelines. 12 The review protocol was registered in the PROSPERO database. 13 The databases of Medline, Scopus Embase, and Cochrane were systematically searched for relevant clinical studies in English by a dedicated librarian without time restrictions in May 2020. The complete search strategy was provided in Appendix 1. The reference list of included studies was crosschecked manually to identify additional studies. All clinical studies reporting patients with primary pancreatic cancer without distant metastasis, who received primary FIO or GEM? regimens, were included. Change in treatment regimen after initial treatment was not an exclusion criterion. Case series with less than ten patients, conference abstracts, letters, reviews, study protocols, and other type of nonoriginal articles were excluded.

Data Collection and Outcome Measures
After removal of duplicates, two reviewers (D.E. and B.A.) independently screened all articles for relevance. Decision of article inclusion in the meta-analysis was done after full-text assessment, and discrepancies between reviewers were resolved by discussion and consensus. The following data were collected for meta-analysis: institution, study design, study population, tumor type (LAPC and BRPC), chemotherapy regimen, radiotherapy application, resection rate, R0 resection rate, and overall and diseasefree survival. If study design was not clearly stated otherwise, the study was classified as a retrospective study. For survival analysis, we obtained patient level data from the studies listed in Table 1. Patients with available survival data from the time of systemic treatment were considered for the ''full treatment population,'' regardless of surgery at later timepoints. Secondary outcomes included resection rates, receipt of radiotherapy, R0 resection rates, and progression-free survival. Progression-free survival was calculated from the start of primary treatment.

Statistical Methods
The individual patient-level data analysis included available time-to-event data from participating centers. Median follow-up time was estimated with the reverse Kaplan-Meier method. Time-to-event outcomes were visualized with the Kaplan Meier method. Hazard ratios for the comparison of FOLFIRINOX versus gemcitabinebased regimens were estimated and reported with 95% confidence intervals (CI). Survival analysis also was reported separately for resected and nonresected patients. The analysis did not account for clustering within centers. A meta-analysis was performed with dichotomous outcomes on study level in single groups, noncomparative data. 14 Due to the expected heterogeneity of included studies, the random effects meta-analysis results were considered most relevant. Single-group, noncomparative data were estimated from all included studies, and proportions were reported either for FOI or GEM? arm. These results were reported with 95% CIs. These results were reported as risk ratios (RR) with 95% CI. The heterogeneity was evaluated according to Cochrane Handbook for Systematic Reviews of Interventions with I 2 statistics. 15 I 2 statistics was interpreted as follows: 0-40% not relevant heterogeneity; 30-60% moderate heterogeneity; 50-90% substantial heterogeneity; 75-100% considerable heterogeneity. Two-sided p-values were calculated in all analyses. The statistical programming language R, version 4.0.3 was used for survival analyses.

Role of the Funding Source
There was no funding for this systematic review with meta-analysis.

RESULTS INCLUDED STUDIES AND DESCRIPTIVE DATA
The systematic literature review identified 317 potential studies, and 10 studies were additionally identified manually. Of those, a total of 23 studies included 2,930 patients in the meta-analysis (Fig. 1). Of these publications, authors of 15 studies provided individual patient-level data for the 2,515 patients 6,10,[16][17][18][19][20][21][22][23][24][25][26][27][28] included in the survival analysis. Among the eight studies without patient level data, one is a randomized, controlled trial, 29 four are prospective, [30][31][32][33] and three retrospective observational studies. [34][35][36] The majority of studies (n = 18) included patients from the start of systemic treatment. Five studies reported only patients undergoing resection. One of the main criteria for study exclusion during full-text assessment was combined reporting of BRPC together with LAPC, or FIO with GEM?. Therefore, separation of data for LAPC and BRPC was possible in all included studies reporting both tumor entities. For the definition of BRPC or LAPC, most studies used NCCN criteria (13 studies; Table 1). In ten studies, either GEM? or FIO was used as systemic treatment; six studies report FIO only, in two of those studies a modified FIO (dose reduction to 75%) regimen was used, and seven studies report only GEM?. Radiotherapy (conventionally fractionated, hypofractionated, or stereotactic) was delivered in 19 studies.

Survival in Resected Patients
In analysis of all resected BRPC patients, patient level data were available for 789 patients with a median followup time of 46.4 months. In patients initially staged as BRPC and resected, median overall survival was 32.9 months (95% CI 28.9-38.2) for FIO, 28. There was no difference in disease-free survival by chemotherapy regimen for resected BRPC patients (p = 0.16, Fig. 3C) or resected LAPC patients (p = 0.75, Fig. 3D

Meta-Analysis of Secondary Endpoints
Secondary analyses of overall resection rate, R0 resection rate, and resection rate after radiation therapy for patients with these reported endpoints: For patients with BRPC, the overall resection rates were 0.53 and 0.55, and R0 resection rates were 0.75 and 0.81 for FIO and GEM? regimens, respectively. In patients with LAPC, overall resection rates were 0.28 and 0.19, and R0 resection rates were 0.72 and 0.71 for FIO and GEM? regimens, respectively (Fig. 4).
Details regarding technique and dose of radiotherapy were lacking from most studies. Thus, we combined chemotherapy regimens and explored the value of radiotherapy in increasing the resection rate in the preoperative setting for BRPC and LAPC. In BRPC, resection rates with and without radiotherapy were 0.58 and 0.51. For LAPC, resection rates with and without radiotherapy were 0.24 and 0.21 ( Supplementary Fig. 1).

DISCUSSION
This study is the largest, systematic review and multiinstitutional, patient-level meta-analysis to examine the impact of neoadjuvant systemic treatment with FOLFIR-INOX (FIO) or gemcitabine-based regimens (GEM?) in patients with BRPC or LAPC. Our data suggest the superiority of FIO over GEM? in both the BRPC and LAPC patient population. However, this benefit is driven in patients who do not undergo surgery. In patients who ultimately undergo resection, the survival outcomes were similar. It may be argued that including patients who received gemcitabine, monotherapy may be considered a palliative treatment and multiagent chemotherapy should be favored whenever possible.
In the setting of primary, resectable pancreatic cancer, surgery followed by adjuvant chemotherapy remains the standard approach. However, neoadjuvant therapy is being increasingly utilized in this setting as evidenced by the increase number of clinical trials. 37 Over the past decades, major advances in novel combination drug therapies and aggressive dose adjustments have significantly improved patient outcomes in the adjuvant setting. Dismal outcomes with 5-FU monotherapy was supplanted by significant survival improvements with the introduction of gemcitabine in 1997. 38 Today, FOLFIRINOX and multidrug, gemcitabine-based regimens have shown benefits in the adjuvant and palliative settings and are now a standard of care. 7,39,40 In contrast to primary, resectable pancreatic cancer, data suggest that patients with BRPC or LAPC have improved outcomes when treated with primary chemotherapy with or without radiotherapy. 29,41,42 Based on the success of FOLFIRINOX and GEM? in the palliative setting, these combination regimens are currently used as preoperative systemic treatment for BRPC and LAPC. There are two available meta-analysis with patient level data examining survival outcomes for FOLFIRINOXbased treatments for BRPC. 43,44 This paucity of data has left significant knowledge gaps for high-level data for other important clinical outcomes (such as resectability and progression-free survival) as well as optimal treatment of patients with LAPC. A recent, multicenter, phase II study from Germany evaluated Gemcitabine/nab-Paclitaxel versus FOLFIRINOX in patients after initial treatment with  gemcitabine/nab-paclitaxel; GemX gemcitabine/cisplatin or oxaliplatin or docetaxel or capecitabine; Gem-mono gemcitabine as monotherapy; ITT full treatment cohort two cycles of Gemcitabine/nab-Paclitaxel and found similar conversion and survival rates in both arms; however, this was not a direct comparison of these regimens. 45 Comparison of the two regimens within two, single-institution, cohort studies did not favor either treatment strategy as both resulted in similar survival rates. 6,8 In patients who do not undergo surgery, our results suggest improved survival rates for patients treated with FOLFIRINOX compared with gemcitabine-based regimens. However, this result was influenced by the outcome of nonresected patients. In patients who underwent resection, we did not observe a benefit for FOLFIRINOX neither in BRPC nor in LAPC. This finding suggests that surgical resection of the tumor remains a critical driver of survival, and the type of chemotherapy may be less important than the duration and sequencing of treatment. Clearly, conversion to surgical resectability also represents the biology of the tumor, implying a less aggressive or chemosensitive tumor. To evaluate outcomes, particularly after surgery, it is critical to stratify by the local stage of the pancreatic tumor. During the past years, the concept of BRPC and LAPC evolved and is currently defined by the tumor involvement of the celiac and mesenteric arteries and the portal vein/superior mesenteric vein. 46 It is important to mention that there is significant heterogeneity in the definitions of BRPC or LAPC across institutions. 47 Moreover, combined reporting of LAPC and BRPC may significantly skew the results, depending on the number of included entities, which was the case in previous meta-analyses. The ability to analyze these patient groups separately is a primary strength of this study. Indeed, the resection rate in two previous meta-analyses differed from 28% to 68% due to the pooling of BRPC and LAPC patients one of these studies. 43,44 The importance of the distinction between BRPC and LAPC is highlighted in this study with the observation that BRPC could be resected in approximately half of the patients after systemic chemotherapy, whereas less than a third of LAPC patients underwent definitive surgical resection after neoadjuvant treatment.
Despite advanced local stage, the R0 resection rates in BRPC and LAPC were high (0.71-0.81) and are well above a recently reported benchmark for pancreatoduodenectomy where R0 resection rates of at least 0.61 were proposed for primary resectable pancreatic cancer. 48 This high R0 resection rate reflects the experience of the included centers and but also may belay the benefit of aggressive neoadjuvant treatment, because these high R0 rates were achieved despite more advanced disease.
Due to a lack of comparative data in our analysis, no definitive conclusion should be drawn regarding the effect of neoadjuvant or definitive radiotherapy, which is similar to limitations discussed in previous meta-analyses. 41,43 There is significant heterogeneity in the treatment fields, biologically effective dose (BED), and fractionation between institutions in treatment of pancreatic cancer. Furthermore, the combination of different modalities in the neoadjuvant setting often precludes current interpretation of a distinct role of one technique over the other. For example, in the recently published, Dutch, PREOPANC-1 study, patients with BRPC were randomized between immediate surgery and primary moderately hypofractionated chemoradiotherapy with Gemcitabine. 29 Preoperative chemoradiotherapy provided a significant overall survival benefit, but the added value of radiotherapy in addition to chemotherapy remains unclear. To this end, the LAP07 trial was unable to demonstrate a survival benefit in the addition of conventionally fractionated, low-BED radiotherapy to gemcitabine monotherapy for LAPC; however, there was a 14% improvement in local control. 49 More recent evidence favors stereotactic radiotherapy as delivered in LAPC-1 after FOLFIRINOX, which demonstrated a survival benefit, but lacks a comparative group not treated with of radiotherapy. 50 Finally, for unresected patients, ablative doses approaching 100 Gy BED may be associated with the best outcomes for LAPC, 51 and new technology, such as MRI-guided, adaptive radiotherapy, may facilitate delivery of this treatment safely. 52 The Alliance for Clinical Oncology Trial A021501compared mFOLFIRINOX with and without hypofractionated radiotherapy in patients with BRPC. The arm with hypofractionated radiotherapy disclosed 67% R1 resection rate, whereas the other arm 43%. This randomized, controlled trial concluded that mFOLFIRINOX alone is superior compared with mFOL-FIRINOX plus hypofractionated radiotherapy in neoadjuvant setting. However, this randomized, controlled trial terminated prematurely and is ultimately underpowered to report the value of the addition of hypofractionated radiotherapy for patients with BRPC. [53][54][55] We acknowledge the limitations of this meta-analysis. There is clearly a lack of prospectively collected, randomized data comparing FOLFIRINOX with gemcitabinebased regimens in patients with BRPC or LAPC. Therefore, the choice of the regimen is left to the physician. 43 None of the included studies reported a decision to use either FOLFIRINOX or GEM?, which reflects the current, real-world dilemma. Due to a more favorable side-effect profile, Gemcitabine-based regimens may be chosen for patients with poor performance status, whereas patients with a higher performance status receive FOLFIRINOX. The number of chemotherapy cycles varied across studies; however, a previous meta-analysis demonstrated this may not have a major impact. 43 Further the data on the rate of complications of systemic therapy, i.e., how many patients interrupt the therapy or need a modification of therapy were not available. Another bias is the role of radiotherapy. Not all included studies performed radiotherapy in the multimodal management of pancreatic cancer in the neoadjuvant setting. The literature search was conducted 18 months ago. Since publications, some further studies were conducted and not included in this analysis. [55][56][57] It is important to highlight that heterogeneity exists between institutional and international committee definitions of BRCP and LAPC. 46 Despite these limitations, our ability to examine patient-level data, examine patients who were and were not resected, and the ability to analyze BRCP and LAPC patients separately offers significant value to this important clinical question. Finally, we were able to offer subanalyses on the variety of combinations of Gemcitabine-based chemotherapy regimens, which further improves the realworld applicability of our findings.

CONCLUSIONS
In the setting of BRPC and LAPC, our data suggest that FOLFIRINOX may be preferred for patients with good performance status due to the survival benefit in patients who are ultimately not resected. Combination therapy with gemcitabine can be considered as a reasonable alternative to FOLFIRINOX, particularly in patients who have a less robust performance status or who are expected to ultimately undergo definitive surgical resection. Prospective data are needed to clarify the optimal timing, duration, and type of chemotherapy regimen as well as the appropriate use of radiotherapy in neoadjuvant or definitive treatment for the heterogeneous patient population that comprises BRPC and LAPC.
AUTHOR CONTRIBUTIONS DE, BA, and RP are first authors, and KL is the last author of this manuscript. UH performed statistical analysis and edited the manuscript. All other authors contributed as following: data interpretation, writing, patient level data. First authors, last author, and a statistician (UH) verified the underlying data.
FUNDING Open access funding provided by University of Zurich. None.
DISCLOSURES Sarah Hoffe: Galera Pharmaceuticals, Institutional Principal Investigator for the GRECO-2 randomized international study for which my cancer center receives funding to support the trial, it is a pancreas trial that includes SBRT. D.E., B.A., R.P share first authorship.
OPEN ACCESS This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/.