Introduction

Pancreatic cancer (PC) is a highly aggressive and refractory disease, with a 5-year survival rate of less than 10%. Recent studies have revealed that pancreatic cancer has become the third leading cause of cancer-related death in the United States [1, 2]. Radical surgical resection remains the only potential cure; however, more than half of patients were not suitable for curative resection due to metastatic or locally advanced disease. Moreover, the survival benefit after resection is more likely to be hampered by the biological behavior of PC [3]. With the development of systematic therapy, there were the compelling evidences from the prospective studies which revealed that neoadjuvant treatment has emerged as an alternative to a surgery-first approach which presented several advantages, including downstaging local disease, systematic control of occult micrometastasis, and increasing the R0 resection rate [4,5,6]. The efficacy and safety of neoadjuvant therapy in borderline resectable and locally advanced PC have been well demonstrated, such that in the newest National Comprehensive Cancer Network (NCCN) guidelines, neoadjuvant treatment was highly recommended for the management of BRPC, as well as high-risk resectable PC [7, 8].

With the growing acceptance of a neoadjuvant treatment strategy for patients with PC, the necessity of precise evaluation of the treatment response has been increasingly emphasized. Although a variety of approaches are used to assess the post-neoadjuvant therapy response, including radiographical changes and changes in tumor marker levels, it is still difficult to estimate the total tumor burden accurately before proceeding with surgery [4, 9]. The heterogeneity of neoadjuvant treatment regimens, inconsistent relationship between tumor response and imaging findings, and lack of a gold standard post-treatment evaluation are the current primary obstacles to the accurate evaluation and treatment of PC patients. Carbohydrate antigen 19-9 (CA19-9) is the most widely used tumor marker in patients with PC. Its utility in determining prognosis and the response to treatment for patients undergoing surgery and chemotherapy has been acknowledged in previous studies. Patients with high levels of CA19-9 at baseline are usually associated with an advanced stage of disease and poor survival [10, 11]. A massive decline or normalization of the CA19-9 level following surgery is a favorable prognostic factor for survival [12]. However, the predictive value of decreasing CA19-9 levels during neoadjuvant treatment for the assessment of the treatment response and survival remains unclear. Numerous studies have reported the beneficial effects of the CA19-9 response after neoadjuvant treatment on survival, but these findings have been limited to retrospective studies in most institutions and there is still insufficient high-level clinical evidence. Therefore, we performed a systematic review and pooled analysis to investigate the association between the change in CA19-9 after neoadjuvant treatment and the prognostic characterization of patients who completed the intended therapy.

Materials and methods

Literature search

A systematic search was undertaken using the PubMed and Embase databases for articles published from March 2010 to March 2020. Two authors independently searched the relevant literature. Combinations of “Pancreatic Cancer”, “Pancreatic Neoplasm”, “Neoadjuvant Therapy”, Neoadjuvant Treatment”, “CA19-9”, and “Carbohydrate Antigen 19-9″ were used as search terms and only articles published in English were included.

Criteria for inclusion and exclusion

Studies eligible for selection had to meet the following inclusion criteria: 1. patients diagnosed with pancreatic cancer were treated with neoadjuvant therapy; and 2. studies with complete information on CA19-9 response and survival for the assessment of hazard ratios (HRs) and their 95% confidence intervals (95% CIs) or provided sufficient data for calculating HRs and 95% CIs from Kaplan–Meier survival curves or using Tierney’s method [13].

The exclusion criteria were as follows: 1. duplicate publications; 2. insufficient data to estimate hazard ratios and 95% CIs; and 3. letters, editorials, case reports, reviews, comments, or meeting abstracts.

Data extraction

Two researchers independently reviewed all selected studies and extracted detailed information, including the first author’s name, publication year, number of patients, resectability of cancer, type of NAT, change in CA19-9 after NAT, number or percentage of surgical patients, follow-up, and survival statistics using a predefined data extraction form. We extracted survival outcome data directly from the studies if the HR and 95% CI were presented or calculated them from the Kaplan–Meier curves using Tierney’s method.

Statistical analysis

Stata software (version 16.0, Stata Corporation, College Station, TX, USA) was used to analyze the extracted data. In this pooled analysis, the prognostic value of the change in CA19-9 after neoadjuvant treatment was evaluated using the pooled HR and corresponding 95% CI. When the HR and its 95% CI did not overlap with 1, the change in CA19-9 after neoadjuvant treatment had a statistically significant effect on survival.

The I2 test was applied to evaluate the interstudy heterogeneity of the HRs. A fixed-effects model (Mantel–Haenszel method) was used when there was minimal heterogeneity among the eligible studies. Otherwise, a random-effects model was applied to calculate the pooled HR when they demonstrated significant heterogeneity. A funnel plot with Begg and Egger tests would be produced to evaluate potential publication bias if sufficient studies were identified.

Quality assessment

Considering that all eligible studies were non-randomized controlled trials, study quality assessment was guided by the Newcastle–Ottawa Scale. Studies with scores ≥ 7 were considered of high quality. Any disagreement was resolved by discussion and consensus agreement.

Results

Study selection and patient characteristics

Retrieved by the aforementioned search strategy, a total of 2478 studies were included initially. After preliminary elimination and secondary further screening, 17 studies met the inclusion and exclusion criteria and were determined accessible for the pooled analysis. These 17 articles, comprised 16 retrospective studies and 1 prospective study. Moderate overall research quality was assessed by the Newcastle–Ottawa Scale with mean score 6.8. The subject number of studies varied from 66 to 419, with a mean size of 131. In total sum, there are 2242 patients receiving neoadjuvant therapy and the therapeutic strategy were various in each institution, including FOLFIRINOX-based systematic chemotherapy, gemcitabine-based chemotherapy (Gemcitabine alone or combined with albumin-bound paclitaxel, S-1), and 5-fluorouracil-based chemotherapy with or without radiotherapy. Preoperative chemotherapy alone was used in 5 studies (Table 1). The literature search process was summarized in a flow diagram based on PRISMA statement (Fig. 1).

Table 1 Characteristics of included studies
Full size table
Fig. 1
figure 1

Flow diagram of the inclusion and exclusion of studies

Full size image

CA19-9 response and survival

The degree of CA19-9 response was characterized in six studies, and among those studies, a decrease of 50% is the most studied change. A pooled analysis showed that CA19-9 response > 50% after neoadjuvant treatment is significantly associated with promising overall survival in PDAC patients (HR = 0.47, 95% CI 0.33–0.61, P < 0.001; I2 = 64.0%, P = 0.02) (Fig. 2).

Fig. 2
figure 2

Overall survival with post-neoadjuvant CA19-9 Response > 50% for patients receiving neoadjuvant therapy

Full size image

A cutoff value of a normal CA19-9 level was widely studied for survival analysis in patients with neoadjuvant treatment and nine studies were included in the following pooled analysis. Successful normalization of CA19-9 after neoadjuvant treatment was significantly associated with favorable overall survival in patients who completed the preoperative treatment and radical surgery (HR = 0.56, 95% CI 0.45–0.67, P < 0.001; I2 = 43.2%, P = 0.08) (Fig. 3).

Fig. 3
figure 3

Overall survival with post-neoadjuvant CA19-9 normalization for patients receiving neoadjuvant therapy

Full size image

Considering both the response > 50% and normalization were optimal indicator of reduction of CA19-9 after neoadjuvant treatment, we did the pooled analysis with this combination. The results showed that optimal CA19-9 response was also promising factors related with prolonged overall survival (HR = 0.49, 95% CI 0.42–0.55, P < 0.001; I2 = 45.1%, P = 0.04) (Fig. 4).

Fig. 4
figure 4

Subgroup analysis of prognostic impact on different method to achieve optimal CA19-9 response

Full size image

Subgroup analyses

We performed three subgroup analyses based on neoadjuvant sequence with or without radiotherapy, chemotherapy duration, and the method of achieving optimal CA19-9 response (CA19-9 > 50% or CA19-9 normalization). No significant deviations from the main results were found in the subgroup analysis of sequence with radiotherapy and the method of achieving optimal CA19-9 response (Figs. 4, 5). Further subgroup analyses regarding the chemotherapy duration showed that patients who received duration of neoadjuvant chemotherapy over 4 months had a survival benefit over those within 4 months (heterogeneity between the groups, P = 0.013) (Fig. 6).

Fig. 5
figure 5

Subgroup analysis of prognostic impact on sequence of radiotherapy in patients with optimal CA19-9 response

Full size image
Fig. 6
figure 6

Subgroup analysis of prognostic impact on duration of neoadjuvant treatment in patients with optimal CA19-9 response

Full size image

Sensitivity analysis and publication bias assessment

Given that heterogeneity was observed in the pooled analysis, a sensitivity analysis was performed for the studies included in this study. Sensitivity analysis after excluding the study by Shimpei et al. showed that the trend supporting the survival benefit of NAT was maintained.

Discussion

To date, surgical resection remains the only curative treatment for PC. However, more than 80% of patients are ineligible for curative surgery and the benefit gained from radical resection of PC is frequently compromised by local tumor recurrence and distant metastases, even in the case of R0 resection. To improve resectability and overall prognosis, neoadjuvant therapy, an emerging therapeutic strategy that has been well-received by patients with breast cancer and colorectal cancer, demonstrated great potential in the systemic control of local disease and occult micrometastasis prior to surgery. Numerous studies have reported significant advantages of neoadjuvant treatment with respect to surgical resectability, R0 resection rate, and overall survival [31]. Restaging evaluation followed by neoadjuvant treatment could identify patients with highly aggressive disease who would not benefit from surgical resection. Imaging examinations and tumor markers are the most commonly used evaluation tools, and CA19-9 is the most commonly used tumor marker in PC. CA19-9 has provided useful information regarding tumor remission and recurrence in operative and metastatic cases; however, it is still unclear how the response of CA19-9 after neoadjuvant treatment would affect the prognosis of patients.

In the present pooled analysis, we found that a CA19-9 response > 50% and normalization of CA19-9 after neoadjuvant treatment were significantly associated with increased OS. Combined analysis revealed that response > 50% and normalization were optimal indicator of CA19-9 reduction to predict the post-neoadjuvant overall survival and potentially guide clinical strategies after treatment. In subgroup analysis, we also found that there was no significant effect on OS whether magnitude of CA19-9 response > 50% or normalization as long as optimal response is achieved. In addition, the duration of neoadjuvant treatment over 4 months was also remarkably associated with favorable prognosis. Previous studies demonstrated that increased or insufficient decrease in CA19-9 usually represented unsatisfactory control of the total tumor burden and probable presence of progressing disease. Research from Takahashi highlighted the poor prognosis in patients with evident chemoresistance and aggressive malignancy [32]. Therefore, concerns about the optimal response of CA19-9 have been raised, when aiming to select appropriate treatment strategies and maximize the patient’s survival. Our work confirmed the use of CA19-9 as a single tool to indicate treatment response after neoadjuvant procedure, and the optimal CA 19-9 reduction would be determined to guide future treatment decisions due to their survival benefit.

Research concerning the magnitude of the proportional change in CA19-9 after neoadjuvant therapy differed between the included studies. Although a reduction of > 50% was the most commonly evaluated, other magnitudes of change were also taken into consideration in the survival analysis. Van et al. reported that a CA19-9 decrease of over 30% was associated with improved survival [33]. Al Abbas et al. indicated that a CA19-9 response of 85% or higher was the optimal threshold for predicting survival and could be incorporated as an endpoint for neoadjuvant therapy [29]. Those studies inspired our further exploration; however, limited as we were to incomplete data, the presented data were not adequate for pooled analysis in this work. Interestingly, we found that the more significant the decrease in CA19-9, the lower the hazard ratio in the survival analysis study, which may lead to the conclusion that patients could benefit more from a greater reduction in CA19-9. However, it is difficult to predict the maximum magnitude of the CA19-9 response during neoadjuvant treatment in clinical practice. Combined our result of impact of long duration of preoperative treatment on CA19-9 response, we hypothesize that maybe prolongation of the chemotherapy cycles or treatment duration to maintain the ideal response level of CA19-9 for a period of time may improve the prognosis of patients [34].

Our work reinforced the CA19-9 as a reliable indicator of the response to NAT in patients with an elevated CA19-9 level at baseline. A massive decrease of > 50% or normalization of the CA19-9 level after neoadjuvant therapy is an effective method for evaluating the treatment response and post-neoadjuvant and postoperative prognosis because the change in CA19-9 is rooted in the tumor biology of PC [35, 36]. Consistent with the research of Hao Liu, the optimal NAT response as evaluated based on the magnitude of the CA19-9 response may predict survival benefit regardless of the regimen [37]. Nonetheless, it may not be suitable for patients who are non-secretors of CA19-9 or have comorbid obstructive cholangitis. For these patients, the change in CA19-9 is disrupted in the restaging evaluation and other markers, such as anatomical changes in radiographic images or inflammatory indices, may be potential replacements. Murthy pointed out that the post-neoadjuvant immune-inflammation index closely paralleled changes in CA19-9 and that its combination with CA19-9 may improve prognostic prediction. Moreover, changes in the standard uptake value on 18fluoro-2-deoxy-d-glucose positron emission tomography–computed tomography could reflect the change in tumor burden and facilitate the indicative function of CA19-9 during neoadjuvant treatment.

To the best of our knowledge, our work is the first systematic review reporting the prognostic value of the CA19-9 response to neoadjuvant treatment; however, there are still some limitations to the current study. First, the majority of analyzed findings were from retrospective studies. A potential imbalance in patient characteristics (performance status and comorbidities) is likely, and findings from more fit patients who potentially received and completed all intended treatment with negative results may exist unpublished. This could be the main reason for the publication bias. Second, heterogeneity may exist in the staging of different tumors resectability and the follow-up duration; therefore, the conclusions may not be generalized to different tumor stages, especially with respect to long-term outcomes. Third, given the insufficient data on the delivered chemotherapy regimens or radiotherapy, the findings did not show the potential effects of different chemotherapy or radiotherapy regimens on survival outcomes between patients with a CA19-9 response. Fourth, the limited available literature suggests that optimal CA 19-9 response after neoadjuvant therapy could be significantly associated with R0 resection rate, negative lymph node metastases, and histopathologic response, but this conclusion cannot be reached due to inadequate data for aggregated analysis in this work.

In summary, serum CA19‑9 values are useful in determining the treatment effect after NAT. The findings of our pooled analysis substantiate the favorable prognostic value of the optimal CA19-9 response in patients with pancreatic cancer receiving neoadjuvant treatment and radical surgery. In the future, the optimal CA19-9 response can be used to guide clinical decision after neoadjuvant therapy according to our research. And we hope the future prospective studies may explore the prognostic impact of long-period maintenance of optimal CA19-9 response through prolonged treatment or more aggressive management.