A critical appraisal of the potential bene ﬁ t of post-operative structured follow-up after resection for biliary tract cancer

Background: There is currently no evidence to support structured use of imaging or biomarkers during follow-up of patients after curative resection of biliary tract cancer (BTC). Besides, the in ﬂ uence of early detection of recurrence and subsequent start of palliative chemotherapy on overall survival remains unknown. The aim of this study is to describe and compare the results of two follow-up strategies. Methods: This retrospective multicenter cohort study compared patients from the Amsterdam UMC undergoing pragmatic clinical follow-up, to patients from the observational cohort of the BILCAP study undergoing structured follow-up. Primary outcome was overall survival. Results: A total of 315 patients were included n=91 pragmatic, n=224 structured follow-up). At median follow-up of 56.9 months, 189 (60%) patients were diagnosed with recurrence. After recurrence, more patients received palliative (chemo) therapy in the structured group (43% vs 75%, P<0.001). Median overall survival was lower in the pragmatic group (27.7 vs 39.1 months, P=0.003). Median overall survival of patients who actually received chemotherapy was comparable (27.2 vs 27.7 months, P=0.574). Conclusion: This study describes the results of two follow-up strategies. Although these groups are biased, it is noted that after pragmatic follow-up fewer patients received palliative chemotherapy but that those who actually received chemotherapy had similar overall survival compared to patients undergoing structured follow-up.


Introduction
Resection, sometimes in combination with adjuvant chemotherapy, is currently the only curative treatment option for biliary tract carcinoma (BTC).Techniques of resection depend on the primary site of BTC and usually entails pancreatoduodenectomy for distal cholangiocarcinoma (dCCA), (extended) hemihepatectomy for intrahepatic (iCCA) and perihilar cholangiocarcinoma (pCCA), and cholecystectomy including gallbladder bed resection for gallbladder cancer (GBC).2][3][4][5][6][7][8] Most common site of recurrence is local regional or in the liver. 3,6,9,10Factors predicting recurrence include poor tumor differentiation, large tumor size, non-radical resection margin, positive lymph node involvement, high tumor stages, perineural growth, and lymph-angioinvasion. [3][4][5]7,[11][12][13][14][15] According to the ESMO and Medscape clinical practice guidelines, follow-up should include quarterly visits with biochemical analysis and radiological imaging for the first two years, not only to identify recurrence but also to detect post-operative complications. 16,17 Hever, these guideline recommendations are not supported by clinical studies or other high level evidence.13 For this reason the Dutch cholangio-and gallbladder carcinoma guideline committee recommends a pragmatic clinical follow-up strategy, and advises not to use radiological imaging or biochemical analysis, including tumor markers, during routine follow-up.18 It is justified to use a pragmatic clinical follow-up given the uncertainty of an improvement of outcome when using structured clinical, biochemical and radiological follow-up.However, most patients with recurrence are often asymptomatic.3,13,19 In these cases structured follow-up may lead to early detection of recurrence and subsequently earlier start of palliative chemotherapy and potentially improve OS.
Structured follow-up has been investigated in studies concerning esophageal and colorectal cancer.1][22][23][24] For colorectal metastasis this is probably due to the local treatment options available for metastatic disease.Unfortunately, clinical and biological features of this disease make extrapolation of these results to BTC impossible.Conversely, for esophageal cancer there are no local treatment options for recurrent or metastatic disease but it still appeared to benefit from structured follow-up with imaging.An important analogy compared to patients with BTC is that when these patients become symptomatic, often because of i.e. biliary or upper gastro-intestinal obstruction, the performance status often quickly deteriorates.This can delay the start of treatment or ultimately lead to the opportunity for palliative chemotherapy being missed completely.
This study aims to describe and compare the results of pragmatic clinical follow-up in a large Hepato-Pancreato-Biliary unit to structured follow-up, using imaging and biomarkers, in a clinical trial in patients with BTC following surgery with curative intent.

Patients
A multicenter, retrospective cohort study was performed including patients with BTC after resection with curative intent.Patients were added from two cohorts, patients with pragmatic follow-up from the Amsterdam UMC and structured follow-up from the observational cohort of the BILCAP study. 19The Institutional Medical Ethics committee of Amsterdam UMC, location VUmc and AMC waived the need for an ethical approval of this study (W21_010, 2021.0359).
The BILCAP study was a phase III study comparing adjuvant capecitabine with observation in patients with BTC following curative intent resection.All patients within the BILCAP study underwent follow-up according to the BILCAP study protocol, including CT scans every 3 months in year 1 and every 6 months in year 2 and annually thereafter.This also included biochemical analysis every 3 months in year 1 and every 6 months in year 2.Only patients from the observational cohort were included in this study.
The Amsterdam UMC cohort, comprises the former VUmc and AMC hospitals, consisted of patients who received a resection for BTC between January 2016 and February 2020 who were identified from the following prospective registries: Dutch Hepato Biliary Audit (DHBA) and the Dutch Pancreatic Cancer Audit (DPCA).Informed consent was obtained using the opt-out method.These patients received pragmatic clinical follow-up according to the Dutch guidelines without routine imaging or biochemical analysis.

Outcomes
Primary outcome was overall survival (OS).Secondary outcome was recurrence free survival (RFS).

Statistical analysis
Data were presented as mean ± standard deviation (SD) or as median and interquartile range when appropriate.OS was defined as months after diagnosis to death or last follow-up using the Kaplan-Meier method.RFS was defined as months after diagnosis to recurrence or lost to follow-up using the Kaplan-Meier method.Survival curves were compared using the log-rank test.The reversed Kaplan-Meier based method was used to calculate median follow-up.Groups were compared using chi-square tests for proportions, Mann-Whitney U tests for medians, and independent sample T tests for means.Time dependent data from the BILCAP study was made comparable to Amsterdam UMC data by adding the time from diagnosis to randomization to all time dependent data points, which were only available in days.Conversion from days to months was based on an averaged conversion factor of 30.4167.Propensity score matching was used to balance baseline differences between the two groups.Variables were selected a priori and included age (<65 year/65 years, lymph node status (N+/N0) and resection margin (R1/R0).Nearest neighbor matching without replacement was performed, with a Caliper width of 0.1.A two-sided P-value of <0.05 was considered to indicate statistical significance.Data were analyzed using IBM SPSS statistics, version 25.0 (IBM Corp).Survival curves were displayed using GraphPad Prism 8 (GraphPad Inc, La Jolla, California, USA).

Baseline characteristics
Of the Amsterdam UMC cohort, a total of 314 patients with BTC met the inclusion criteria.From these, 223 patients were excluded because of study participation with structured follow-up, leaving 91 patients for inclusion.A flow diagram of the Amsterdam UMC patient selection is presented in Supplementary Fig. 1.
Of the BILCAP study cohort, a total of 224 patients from the observational arm were included.This resulted in a total group of 315 patients.Of these patients 162 (51%) were male.The median age at diagnosis was 65 (55-70) years.Distal cholangiocarcinoma was the most common type of BTC (n = 118, 38%) and pancreatoduodenectomy was the most performed type of surgery (n = 116, 37%).Time from diagnosis to surgery was 44 (32-68) days.In 198 (63%) patients the resection margin was negative at final pathology report.Lymph nodes were negative in 158 (50%) patients and the majority of the patients had a moderately differentiated tumor (n = 171, 54%).All baseline characteristic are displayed in Table 1.

Propensity score matching
After propensity score matching, two groups of 85 patients were obtained.After matching, age, differentiation grade, tumor size and first post-operative CA19.9 measurment were comparable between both groups.Only ECOG performance score, resection type and first post-operative bilirubin levels were significantly different between both groups.All baseline characteristic after matching are also displayed in Table 1.

Recurrence
In the total study cohort of 315 patients, 189 (60%) were diagnosed with recurrence during follow-up, 46 patients (51%) in the pragmatic follow-up group and 143 patients (64%) in the structured follow-up group, P = 0.029.In the propensity matched cohort, 95 (56%) out of the 170 patients developed recurrence, 45 patients (53%) in the pragmatic follow-up group and 50 patients (59%) in the structured follow-up group, P = 0.440.
Local recurrence was found in 82 patients (43%).Distant recurrence was found in 102 patients (54%).An increase in CA19.9 level, with a median level of 126 U/ml (44-239) and a median accompanied low bilirubin level of 9 mmol/L (6-13.5),indicated the first sign of recurrence in 73 (39%) of the 189 patients.This included 4 patients (9%) in the pragmatic follow-up group and 69 patients (48%) in the structured follow-up group.In 61 patients (32%) recurrence was first suspected on imaging.This included 8 patients (17%) in the pragmatic follow-up group and 53 patients (37%) in the structured follow-up group.In 34 patients from the pragmatic follow-up group (18%), the first sign of recurrence were clinical complaints (e.g.pain, cholangitis, weight loss), Table 2.
In the group without recurrence (n = 126), 25 (20%) patients showed an increase of their CA19.9levels at one point during follow-up.In one patient this was related to an increase of the bilirubin level.In 20 out of the 25 patients a CT scan was perfomed, which was negative for recurrent disease in all patients.

Overall survival
Of the 315 patients, 186 had died within the follow-up period.Median follow-up of patients alive at last medical visit was 56.9 months (95% CI 48.2-65.6).
From the 189 patients with recurrence, 127 (67%) patients were treated with palliative (chemo) therapy.This included 20 patients (43%) in the pragmatic follow-up group and 107 patients (75%) in the structured follow-up group, P < 0.001.From the total group, 59 patients did not receive palliative (chemo) therapy.This included 23 patients (50%) in the pragmatic follow-up and 36 patients (25%) in the structured follow-up group.For three patients this information was unknown.Impaired performance status was the reason for not receiving palliative (chemo) therapy in 18 (78%) out of the 23 patients of the pragmatic follow-up group, for patients with structured follow-up this was unknown.
In the propensity matched cohort, 53 (56%) out of the 95 patients were treated with palliative (chemo) therapy.This included 19 patients (42%) in the pragmatic follow-up group and 34 (68%) patients in the structured follow-up group.

Discussion
This study is the first retrospective cohort study comparing two follow-up strategies in patients with resected BTC and found that patients undergoing structured follow-up have a favorable OS compared to patients with pragmatic follow-up (39.1 months  structured vs 27.7 months pragmatic).After propensity score matching, this observation persisted.Although this comparison is not perfect, given the obvious bias and selection issues, the findings provide some interesting observations.In total, 189 (60%) patients were diagnosed with recurrence of whom 127 where treated with palliative (chemo) therapy.Interestingly, this included 107 (75%) patients from the structured and only 20 (43%) patients from the pragmatic follow-up group.Median OS of patients who actually received palliative (chemo) therapy was comparable in both groups, this remained after propensity score matching.These findings could lead to the question if perhaps early detection of recurrence is important for the OS in patients with BTC.Early detection of recurrence could result in a timely start of chemotherapy maybe resulting in increased OS.Clinical detection of recurrence often includes jaundice, cholangitis and peritoneal dissemination with ascites and/or bowel obstruction.These challenging clinical situations frequently occur in individuals who are already vulnerable.Perhaps structured follow-up leads to early detection and start of palliative (chemo) therapy before clinical performance deteriorates.
Recently, an international multicenter study exploring the impact of intensive structured follow-up after esophageal cancer surgery on oncological and quality of life outcomes was for the structured follow-up group (P = 0.393) Figure 2 (Left): overall survival displayed for both groups.The median and 5 year OS divided for follow-up method was 27.7 months (95% CI 22.0-33.4)and 14% for the pragmatic follow-up group and 39.1 months (95% CI 32.7-45.4)and 36% for the structured follow-up group (P = 0.003).# For two patients time to diagnosis and therefore total time from diagnosis to death or last follow-up was not available.(right): Overall survival of the propensity matched cohort.The median and 5 year OS divided for follow-up method was 6.3 months (95% CI 21.9-30.6)and 13% for the pragmatic follow-up group and 40.8 months (95% CI 27.1-54.5)and 43% for the structured follow-up group (P = 0.003).*For one patient time to diagnosis and therefore total time from diagnosis to death or last follow-up was not available published.Results of the ENSURE study (NCT03461341), showed that structured follow-up was associated with reduced symptomatic recurrence and increased tumor-directed therapy, with a clear OS benefit. 20he difference in OS between both groups in the current study cannot easily be explained by differences in tumor biology, as the tumor stage (T), lymph node stage (N), and differentiation grade are comparable between groups and in the propensity matched cohorts.The only difference between groups was the high ECOG score in the pragmatic follow-up group which may have had an effect on the outcome.Besides it is known that participation in a clinical trial improves patient outcomes, which can mostly be attributed to indirect positive effects. 28,29t can be argued that pragmatic follow-up has several clear advantages compared to structured follow-up.As mentioned before, patients with recurrence are most often asymptomatic. 3,13,19Many oncologists will be reluctant to give palliative chemotherapy to asymptomatic patients, simply because there are no complaints to palliate and chemotherapy will merely decrease quality of life (QOL) with only a modest benefit for OS. 30 This makes structured follow-up futile as long as there are no complaints amenable to palliative chemotherapy or any proof that starting early may improve OS or QOL.
Currently, imaging using Computed Tomography (CT) may be the most reliable follow-up method as biochemical followup using Carbohydrate Antigen 19.9 (CA19.9), a serum tumor marker, 31 is subjected to several limitations.Patients may experience elevated CA19.9 levels without signs of recurrence and about 10% of patients are unable to synthesize CA19.9 resulting in potential false negative outcomes. 32Results can also be false positive for other reasons, for example an increase of CA19.9 can be the result of cholestasis. 1,33 -35Additionally, due to low disease activity and difficult reachable locations, histological confirmation of recurrence can be challanging.This causes a delay in confirmation of recurrence and consequently start of palliative chemotherapy.
This study was not able to point to the best method of followup.Frequent measuring of CA19.9 seems straightforward and inexpensive.However, as shown in this study, high CA19.9levels do not always accurately indicate recurrence.In the group without recurrence, 25 patients showed an increased CA19.9 level at one point during follow-up.This led to a negative CT scan in 20 patients.It is unknown why some patients have elevated levels without other signs of recurrence. 34he findings of this study should be interpreted with caution due to the fact that this is a retrospective cohort study including patients from a clinical trial and patients who are treated with pragmatic clinical practice.In an attempt to compensate for bias and selection, a propensity score matched analysis was performed.Nevertheless, selection bias has probably influenced the composition of the included groups.For the Amsterdam UMC group for example, only 91 patients could be included.Most of the excluded patients participated in the ACTICCA-1 trial.Patients who were not included in the ACTICCA-1 study either waived the study or may have been excluded because of to an impaired post-operative functional status.On the other hand, this study is based on the only available data simply because there have been no trials or prospective studies on this subject for BTC.
This study shows several interesting observations which can be used in a future randomized controlled trial comparing pragmatic and structured follow-up.Looking slightly further, it could be interesting to stratify patients in this future trial for positive lymph nodes.It is known from earlier studies that patients with positive lymph nodes more often show early recurrence. 10,36This might lead to a different strategy for those patients.Besides, as mentioned earlier, CA19.9 is not always the best marker indicating recurrence.Currently there is a growing interest in Circulating Tumor DNA, which could offer an interesting addition to the clinical pathway for detecting recurrence. 37n conclusion, this study describes two commonly used strategies for follow-up of patients with BTC following resection with curative intent.Conclusions whether one strategy is superior to the other cannot be drawn as result of the bias in the selected groups.Nevertheless several interesting observations are made which may help to set up a future randomized controlled trial comparing pragmatic and structured follow-up.

Table 1
Baseline characteristics before and after matching.

Table 1
(continued ) P-values based on complete case analysis unless unknown is displayed b First measurement post-operative.c Only available in 74/224 patients of the BILCAP data. a

Table 2
Follow-up characteristics a P-values based on complete case analysis unless unknown is displayed.b Combination of surgery/radiotherapy/chemotherapy.c If CA19.9 was the first sign of recurrence.