The importance of time‐to‐adjuvant treatment on survival with pancreatic cancer: A systematic review and meta‐analysis

Abstract Background While adjuvant chemotherapy benefits patients with pancreatic ductal adenocarcinoma (PDAC), the importance of the time to initiation of adjuvant therapy remains unclear. Aim This study seeks to better understand whether the timing of postoperative chemotherapy initiation affects long‐term outcomes in PDAC. Methods and Results A systematic literature search was performed in Medline, Embase, and Cochrane Library in March 2020. Studies focused on the association between the timing of adjuvant therapy on long‐term outcomes in resected PDAC patients were included. The impact of early and delayed therapy as defined by the respective studies was evaluated using forest plot analysis. Overall survival (OS) and disease‐free survival (DFS) served as primary endpoints. Out of 3099 published articles, 10 retrospective studies met inclusion criteria. Combined, these studies included clinical data of 13 344 patients. The cut off used to define “early” and “delayed” treatment groups varied in the included studies ranging from 3 to 12 weeks. Due to this heterogeneity, a sub‐group analysis of three time cut offs was performed: 3 to 5 weeks, 6 to 8 weeks, and 9 to 12 weeks. There was a significant decrease in OS and DFS when adjuvant therapy was delayed by 3 to 5 weeks after surgery (OS, pooled hazard ratio [HR] = 1.86, 95% confidence interval [CI] = 1.25‐2.78; DFS, pooled HR = 1.62, 95% CI = 1.12‐2.34). However, due to small sample size and limited studies in this subgroup analysis, the results may be indeterminate. There was no significant decrease in OS with delayed initiation of adjuvant therapy by 6 to 8 weeks and 9 to 12 weeks. Similarly, delay in adjuvant therapy beyond 3‐5 weeks. Conclusions There was no conclusive evidence suggesting improved survival in patients starting treatment at various time cut offs. Studies investigating the extreme ends of the time‐to‐treatment spectrum may prove more informative.


| INTRODUCTION
Pancreatic ductal adenocarcinoma (PDAC) is the most lethal common cancer. The overall five-year survival remains just 10%. 1,2 Most patients present with unresectable disease. However, approximately 20% of patients have localized and resectable disease. 3 The best chance of long-term survival for such patients is resection combined with multi-agent chemotherapy. 4 Historically, chemotherapy was not considered as decidedly beneficial following surgical resection until the late 1990s. 5 Based on the results of the european study group for pancreatic cancer (ESPAC-1) trial in 2004, 6 adjuvant chemotherapy using single agent 5-fluorouracil (5-FU) became the standard of care for resectable PDAC. A series of trials showing equipoise between gemcitabine and 5-FU but a better toxicity profile in the adjuvant setting for gemcitabine shifted preference to this drug. [7][8][9] Recent trials have determined a survival benefit for multi-agent chemotherapy regimens when compared with single agent therapy. 5 Despite the benefit of adjuvant chemotherapy, nearly half of the patients fail to receive additional therapy following resection, often due to a complicated postoperative course. 10 Indeed, the time from a patients' operation until starting adjuvant therapy can vary widely. [11][12][13][14][15][16][17][18][19][20][21][22] Some patients may never recover enough to receive any amount of adjuvant treatment. Alternatively, patients may not tolerate multi-agent chemotherapy after resection and instead receive less effective single agent therapy. The optimal timing for initiating adjuvant chemotherapy has not been rigorously evaluated. Most versed in the literature refer to a single post-hoc analysis of a randomized trial which showed that adjuvant chemotherapy can be safely started up to 12 weeks post-surgery. 12 While there appears to be a benefit of adjuvant chemotherapy based on randomized adjuvant trial data, the impact of the timing of treatment in the general population is a question that requires further evaluation. Herein, we performed a meta-analysis of the available literature to characterize the time from surgery to initiation of adjuvant therapy and determine the effect of delay on overall survival (OS) and disease-free survival (DFS).

| Protocol and registration
This systematic review and meta-analysis was conducted following the recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-analyses statement (PRISMA). 23

| Data extraction and synthesis
The following data points were extracted: sample size, patient demographics, time cut-offs delineating early and delayed initiation of adjuvant therapy groups, chemotherapy regimen used, median follow-up period, usage of univariate, or multivariate survival analyses, and comparisons of median OS or DFS endpoints. The Newcastle-Ottawa Scale (NOS) was used to ascertain the quality of observational studies and determine risk of bias. 24 NOS scores ranged from 0 to 9 and can be categorized into three groups: very high risk (0-3), high risk, 4-6 and low risk of bias. [7][8][9] Hazard ratio (HR) was used as the measure of effect for comparisons of OS and DFS between delayed and early treatment groups.
Delayed chemotherapy was defined as chemotherapy started beyond a certain cutoff time period as described in the respective studies. For each study, the adjusted HR and 95% confidence interval (CI) were annotated, and SE was calculated from available data. If multivariable analyses were not performed, univariate HR was recorded. For studies where HR was not recorded, Kaplan-Meir curves were digitalized using Webplot digitalizer software 25 and HR and 95% CI was estimated. 26 For both OS and DFS, each publication was weighted as a function of the inverse variance of each effect size and forest plots were constructed. Cochrane Chi 2 and I 2 statistics were used to assess homogeneity for each outcome. Studies were considered to have significant heterogenicity when Chi 2 P-value was less than .1 and I 2 was greater than 50%.
The pooled HR for OS and DFS with early vs delayed adjuvant therapy was calculated either using the fixed effects model/Manzel-Haenzel method or random effects model/DerSimonian-Laird method based on heterogeneity of the included studies. Random effects method was used when I 2 was greater than 50%.
The publication bias was evaluated using Egger's linear regression and funnel plot analysis to illustrate asymmetry between studies.
Studies were considered to have significant publication bias if P-value was less than .05. The GRADE approach was utilized to evaluate the quality of evidence of this meta-analysis. The assessment includes risk of bias, imprecision, inconsistency, indirectness, publication bias, magnitude of effects, dose-response relations, and impact of residual confounding and bias. 27 Using the above parameters, the GRADE certainty rating is graded as very low, low, moderate, and high. The

| RESULTS
The literature search identified 3099 studies ( Figure 1). After screening of titles and abstracts, complete text of 13 studies were obtained for further review. Of the 13 studies, two were excluded as they did not study the effect of timing of adjuvant therapy on long-term outcomes. In total, we included 11 publications for descriptive analysis in this study. There were no non-English studies on the subject topic. Of these 11 studies, one utilized three time periods and was excluded.
Ten studies utilized two time periods: early and delayed-treatment groups. These 10 studies were further used for performing the metaanalysis. Table 1

| Timing of adjuvant chemotherapy on OS
Ten studies evaluated the effect of a delay in initiating adjuvant chemotherapy on OS. In total, these studies included 13 344 patients.
The number of patients was nearly equally distributed between early and delayed-treatment groups.
Due to the varying time cut-offs used in the 10 studies to define early and late treatment groups, subgroup analyses were performed for the following cut-offs: 3 to 5 weeks, 6 to 8 weeks, and 9 to 12 weeks. Given the heterogeneity in the definition of the study groups, random effects model was chosen for all further analyses.
There was a significant decrease in OS in the two studies where adju-  I, II, III, IV   I, II, III, IV   I, II, III   I, II, III   I, II, III   I, II, III   NR   I, II, III   I, II, III   I, II   I, II,

| Publication bias
The funnel plot was used to study the degree of asymmetry of individual study results around the pooled HR for OS ( Figure 5A). Asymmetry was detected which was further analyzed using the Egger method. There was significant asymmetry in study results (P = .02) which indicates a publication bias. A funnel plot was also constructed for DFS and was found to be symmetrical ( Figure 5B), as confirmed by Egger test (P = .59).

| GRADE certainty rating
The GRADE certainty assessment is shown in Table 3. The GRADE rating indicated very low and low quality of evidence for the metaanalyses evaluating OS and DFS, respectively.

| DISCUSSION
The role of adjuvant chemotherapy for PDAC has been extensively studied over the past two decades. As the majority of PDAC are usually associated with micrometastases at presentation, even at the clinically resectable and localized stage, 28   Two additional reports from 2016 found no differences in OS when comparing those who started chemotherapy within 8 weeks of surgery and those who started treatment after the eight-week timepoint. 14,15 Lee et al 17 reported that early vs late initiation of adjuvant therapy, defined as before or after 6 weeks postoperatively, did not impact OS. However, the authors also commented that patients who were able to complete therapy had a significant survival advantage. A multiinstitutional study from 2017 showed that timing of adjuvant therapy (before or after 12 weeks postoperatively) did not affect survival, while those who received surgery alone had a reduced OS compared with patients who received adjuvant chemotherapy at any time. 18 Finally, a recent meta-analysis showed no differences in survival when comparing patients who started chemotherapy within 6 to 8 weeks to those who started after 8 weeks. 20 The aforementioned meta-analysis differs from the present one in that it was more limited in scope. Studies evaluating TTT with cut offs other than 6 to 8 weeks were not included. The study only included six studies in their analysis (vs 10 here). Also, DFS was not considered as an outcome. A thorough risk of bias analysis and effect on the certainty of evidence were not investigated.
Contrary to the previously mentioned papers, a few studies did find an association between timing of chemotherapy and survival. For instance, Kim et al 16 reported that patients who received treatment within 5 weeks of surgery had significantly better OS, as compared to those who started therapy after this cutoff. However, patients who were not able to complete their adjuvant treatment regimen faired significantly worse. A large study using the NCDB showed that early adjuvant chemotherapy (started within 4 weeks of surgery) and delayed initiation (started after 8.4 weeks) had higher mortality rates compared to those who started chemotherapy between 4 and 8.4 weeks. 19 This study on its own might suggest that initiation too early may put the patient at risk, and that a "sweet-spot" timeframe was possible. White et al 21 performed a propensity score matched analysis of NCDB patients and concluded that patients who received chemotherapy before 66 days had a better survival advantage.
From our meta-analysis, patients who received adjuvant therapy 3 to 5 weeks after surgery had decreased DFS and OS. However, this sub-group analysis suffers from the limitation of small sample size derived from just two retrospective cohort studies. Also, our results demonstrate a wide confidence interval, which could suggest a relatively imprecise estimate of outcome. We found no significant difference in DFS and OS between "early" and "delayed" adjuvant therapy groups at 6 to 8 week and 9 to 12-week cut-offs.
Despite the fact that all of the included studies were observa- It is also important to recognize that modern chemotherapy, albeit beneficial, only offers a marginal survival advantage. OS improvements are typically between 2 and 4 months in most randomized trials, while DFS is around 7 months. 37 The recent PRODIGE trial testing FOLFIRNOX is a noteworthy exception with a much stronger reported benefit. 38  The study has technical limitations that also deserve mention.
The included studies had varying definitions for "early" and "delayed" treatment groups. Though we performed subgroup analyses and meta-regression to account for the significant heterogeneity of studies, certain subgroups consisted of statistically less complex or smaller studies, which may not provide accurate estimates of the actual effects. Certain studies reported adjusted HR whereas others had unadjusted HR, which could lead to bias on the pooled analysis. In the studies that did not provide HR, there may be a possibility of error in synthesizing the HR from the Kaplan Meier curves. Also, this systematic review and meta-analysis is prone to potential errors in search methodology, selection, and reporting bias. Given the paucity of prospective research in this subject topic, the certainty of evidence of this meta-analysis is low. An ideal study would involve a prospective, multi-center, randomized clinical trial. However, feasibility of such a prospective study would be questionable, provided that ethical concerns in regard to delaying adjuvant therapy in this subset of patients.

| CONCLUSION
Our meta-analysis shows that there was no conclusive evidence suggesting improved survival in patients starting treatment at various time cut offs. Given the paucity of prospective studies, the results need to be cautiously interpreted. Further multi-institutional studies utilizing similar chemotherapy regimens that compare the extreme ends of the treatment spectrum are required. Based on our understanding of the natural history and biology of PDAC, time-totreatment should be optimized with a goal to deliver treatment as soon as the patient is clinically recovered from surgery and considered to be fit enough to tolerate chemotherapy.

CONFLICT OF INTEREST
The authors declare no conflicts of interest.

ETHICAL STATEMENT
This study was approved and registered as a systematic review in the PROSPERO database on 04/28/2020 (Registration number: CRD42020170486).

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.