Analysis of the CONKO-001 trial: Is the validity of the study sufficient to recommend adjuvant chemotherapy for pancreatic cancer?

Abstract

BACKGROUND:

Since randomized controlled trials have indicated that adjuvant chemotherapy prolongs survival and reduces recurrence rates after surgical resection of pancreatic adenocarcinoma, a gemcitabine based chemotherapy has become part of the interdisciplinary treatment concept for pancreatic cancer in accordance to current guidelines.

OBJECTIVES:

The aim of this project was to analyse the validity of the CONKO-001 trial as a basis for the recommendation of adjuvant chemotherapy in many international guidelines.

METHODS:

We analysed the validity of the CONKO-001 trial regarding study design, recruitment period, participating institutions, patient selection, randomisation, stratification, standardization of surgical treatment and histological examination, statistical methods and interpretation of results. We additionally analysed the study regarding the risk of bias using the RoB 2 Tool. Finally we reviewed the influence of the pharmaceutical industry and potential conflicts of interest.

RESULTS:

We identified several shortcomings of the study concerning the study protocol, the participating clinics, the patient recruitment, the randomization pattern, the standardization of surgical treatment and histological examination, the statistical methods, the evaluation of the results and the influence of the pharmaceutical industry. According to the Cochrane RoB 2 Tool the study was judged to raise some concerns in three of the five risk domains for the outcome “overall survival”.

CONCLUSIONS:

Based on our review, the results of the CONKO-001-study should be revisited and critically reviewed. The recommendation to include adjuvant chemotherapy with gemcitabine deserves a critical appraisal.

Keywords

Pancreatic cancer adjuvant chemotherapy validity adenocarcinoma guidelines recommendation randomized controlled trials

1. Background

One of the primary objectives of any medical therapy is to avoid harming the patient by choosing wisely what kind of treatment would be the most promising and what risks are necessary to be taken. Those risks of therapy should be discussed particularly if the treatment period is short and the patient’s life time is limited due to malignant disease. Pancreatic adenocarcinoma still has a very poor prognosis, while incidence is increasing. Primary resection is possible only in a minority of patients. Diverse efforts have been made to prolong the patients survival after resection, including neoadjuvant and adjuvant radiotherapy, chemotherapy, and radiochemotherapy. Since the 5-year-survival rates rarely exceed 20% even after curative resection, international guidelines have recommended adjuvant chemotherapy after resection of pancreatic cancer, regardless of tumor size and lymph-node status. Following initial 5-FU-based chemotherapy regimens, the administration of gemcitabine became the established therapy for a long time on the basis of lower toxicity with the same effectiveness [1]. Presumably more effective chemotherapy regimens with higher toxicity rates were adopted in younger patients with a generally higher performance status.

The ESPAC-1 trial was one of the first to relevantly influence the chemotherapeutic regimen. According to the study, which included 256 patients, Neoptolomos et al. [2] concluded that 5-FU-based chemotherapy can significantly extend the overall survival of patients with resectable pancreatic cancer, whereas radio(chemo)therapy was more likely to have negative results. This however, was investigated in a study with an untransparent two-factorial design.

We already analyzed this study in detail and concluded that adjuvant 5-FU-based chemotherapy after pancreatic resection with curative intention cannot be recommended on the basis of the ESPAC-1 study [3]. However, this study serves as a reference in many guidelines [4–8]. The goal of our current work is to examine the validity of a further important study on this topic. After reviewing the guidelines which recommend adjuvant chemotherapy after pancreatic resection, we chose the CONKO-001 study [9,10] for our validity analysis because it is cited in many international guidelines. For example the German S-3 guideline (2013) [4], the ESMO guideline (9/2015) [5], the NCCN guideline (11/2019) [6], the SEOM guideline (2016) [7] and the SCAN guideline (2015) [8] recommend administration of adjuvant gemcitabine after surgery of resectable pancreatic carcinomas on the basis of the results from the CONKO-001 study. This study [9,10] was designed as a multicenter, open-label, parallel-group study with centralized randomisation to investigate whether adjuvant chemotherapy with gemcitabine could influence the survival of patients with pancreatic cancer after resection. An improved 5-year-survival of 11% after administration of adjuvant gemcitabine was reported.

2. Materials and methods

The interim results of the CONKO-001 study were published by Oettle et al. in JAMA [9], whereas the long-term results were published by the same group [10]. We analyzed both publications in terms of study design, recruitment period, patient selection, randomization, participating centers, surgical quality, statistical methods and interpretation of results.

We also analyzed the long-term study [10] regarding the risk of bias using the Cochrane RoB 2 Risk of Bias Tool. This tool was first introduced in 2008 [11] and revised in August 2019 [12]. The RoB 2 tool determines the risk of bias in a single estimate of intervention effect for a single outcome or endpoint rather than for a whole trial as the risk of bias is outcome specific. Bias is structured into five distinct domains: (1) bias arising from the randomization process, (2) bias due to deviations from intended interventions, (3) bias due to missing outcome data, (4) bias in measurement of the outcome and (5) bias in selection of the reported result. The risk-of-bias judgements are based on the responses to the signaling questions. Their response options are: “yes”, “probably yes”, “probably no”, “no”, and “no information”. According to the given answers it is possible to judge the risk for each domain. The risk of bias judgments for each domain are “low risk of bias”, “some concerns”, or “high risk of bias”. Thereafter, an overall risk-of-bias judgment can be provided for the study for a specific result (e.g. overall survival) as follows: low risk of bias if the study is judged to be at low risk of bias for all domains for this result, some concerns if the study is judged to raise some concerns in at least one domain for this result, but not to be at high risk of bias for any domain and high risk of bias if the study is judged to be at high risk of bias in at least one domain for this result, or the study is judged to have some concerns for multiple domains in a way that substantially lowers confidence in the result [12].

Finally we elucidated the influence of the pharmaceutical industry and conflicts of interest.

3. Results of the analysis of both CONKO-001 study publications

3.1. Study design, participating institutions and patient recruitment

The trial was initiated by the German Cancer Society and coordinated by the Charité- University Clinic in Berlin. 88 academic and community-based oncology centres in Germany and Austria participated. Patients with histologically proven, macroscopically completely resected pancreatic cancer without distant metastases regardless of the tumour size and nodal stage were included.

The original protocol included only patients with confirmed R0-resection, defined as histologically tumor-free surgical margins. In November 1998, after the recruitment of 9 patients, the protocol was changed to also include patients with R1 resection because this offered the possibility of performing separate end point analyses in patients with R0 and R1 resection to compare the effect of adjuvant gemcitabine in these subgroups. The study protocol did not require a standardized system of pathologic assessment or pathologic quality control.

The primary endpoint was disease-free survival (time from randomization to first documentation of recurrence); secondary endpoints were treatment safety and overall survival (time from randomization to death or last follow-up). In the first publication quality of life was also defined as a secondary endpoint, measured with the Spitzer Quality-of-Life Index, while in the second publication quality of life was not mentioned any more.

3.2. Randomisation and stratification

Randomisation into the treatment group (gemcitabine) or the control group (observation) was stratified according to tumor stage (T1-2 versus T3-4), nodal status (N0 versus N1) and resection margins (R0 versus R1).

Using scaled envelopes, randomization was conducted in the first 73 patients by a statistician at the German Cancer Research Center, while randomization of the remaining patients was performed at the coordinating center in Berlin using a computer-generated procedure. A reason for this modification of the randomisation procedure was not given.

3.3. Standardization of surgical treatment and histological examination

Surgery was performed according to institutional standards, and the histological examination was performed by the pathology institutes of the recruiting centers without central review.

3.4. Treatment (chemotherapy)

Adjuvant chemotherapy could be administered either by oncology departments in hospitals or on an outpatient basis by oncologists. The patients received six cycles of gemcitabine every four weeks consisting of three weekly infusions of gemcitabine (1000 mg/m²) followed by a one-week-break.

Patients were withdrawn from the study for any of the following reasons: recurrent disease, wish of the patient, unacceptable toxicity, pregnancy or inadequate contraception in a woman of childbearing potential, or if the patient was likely to benefit more from an alternative treatment according to the investigators discretion.

Patients in the observation group who experienced recurrent disease were offered palliative chemotherapy, generally with gemcitabine.

Patients received follow-up every eight weeks for up to five years or until death.

All patients had given written informed consent before randomization.

3.5. Statistical methods

Assuming that median disease-free survival after complete resection of pancreatic cancer without adjuvant treatment would be less than 12 months, a number of 184 patients in each group was calculated to be required to detect an improvement in six-month disease-free survival by the administration of gemcitabine. This calculation was performed with a bilateral alpha error of 5% and a power of 90% assuming a patient dropout of 20%. If the calculation could also be applied for the secondary endpoint of overall survival is not described. A recruitment period of three years with an additional follow-up period of two years was planned. The established tests (Kaplan Meier survival curves, Cox model, log rank tests, SPSS) were used for statistical analyses, which were carried out according to intention-to-treat. Patients lost to follow-up were censored on the date of last follow-up.

4. Recruitment period and results of CONKO-001

The CONKO-001 study recruited 368 patients from 88 institutions in Germany and Austria over a period of 6.5 years (1998 to 2004). So the planned recruitment time was more than doubled. From the 368 recruited patients 179 patients in the Gemcitabine group and 175 patients in the observation group met the eligibility criteria and constituted the intent-to treat-population for the primary end point analysis. All but 11 patients had adenocarcinoma. The baseline characteristics of the two treatment groups were similar. Most of the patients had T3 tumors, a G2-grading and a positive lymph node status (N1). The exact TNM-stadium was not given. 62% of the patients randomized to adjuvant treatment received all six cycles of gemcitabine.

The analysis of 2007 was done after a median follow-up of 53 months (range 9–96). The median disease-free survival was 13.4 months in the gemcitabine group compared to 6.9 months in the observation group. This result was significant (p < 0.001). Subgroup analysis demonstrated that the beneficial effect of Gemcitabine was evident for all stratified groups (R0/R1, T1-2/T3-4, N0/N1). All the analyses were performed according to the intent-to-treat protocol but also according to qualified analysis, which delivered similar results.

At the time of the analysis, there was no significant difference in overall survival (median 22.1 months for the Gemcitabine group and 20.2 months for the observation group, p = 0.06).

The Kaplan–Meier curves in the primary study are shown in Fig. 1.

Fig. 1.

Kaplan–Meier curves of disease free-survival (A) and overall survival (B) of the treatment groups of the CONKO-001-study [9] (interim results).

After closing the database in September 2012, the long-term results (published 2013) showed a significant increase in disease-free survival for patients treated with gemcitabine for six months after resection of pancreatic cancer (median: 13.4 months) compared to the observation group (median: 6.7 months). This treatment effect was detected across all the stratified subgroups. The median follow-up duration was 136 months (range 104–144).

The median overall survival also increased significantly (p < 0.01) from a median of 20.2 months (observation group) to 22.8 months (treatment group). These results are very similar to the non-significant results of the interim-analysis (20.2 months for the observation group and 22.1 months for the gemcitabine group). Survival rates show equivalent results: after five and ten years respectively, the rates of disease-free survival were 16.6% and 14.3% in the gemcitabine group versus 7.0% and 5.8% in the observation group; the rates for overall survival were 20.7% and 12.2% in the treatment group and 10.4% and 7.7% in the observation group. All results are shown in Table 1. The treatment effect of gemcitabine based on the stratification criteria was seen in all subgroups. The Kaplan–Meier curves are shown in Fig. 2.

Table 1

Median survival and survival rates of the treatment groups of the CONKO-001-study [10] according to adjuvant treatment with gemcitabine or observation

	Observation group	Treatment group
Median disease free survival	6.7 months	13.4 months
Median overall survival	20.2 months	22.8 months
Disease free survival at 5 years	7.0%	16.6%
Disease free survival at 10 years	5.8%	14.3%
Overall survival at 5 years	10.4%	20.7%
Overall survival at 10 years	7.7%	12.2%

Fig. 2.

Kaplan–Meier curves of disease free-survival (A) and overall survival (B) of the treatment groups of the CONKO-001-study [10] (final results).

In comparison with the Kaplan–Meier curve of the interim analysis (Fig. 1) there are different numbers of patients at risk, although the curves start with the same number of patients. In the Gemcitabine group of the interims analysis 73 patients were still alive in contrary to 87 patients in the long-term study at this time.

5. Conflict of interest

The authors Oettle and Riess received financial remuneration for contributions to advisory boards or oral presentations from Roche Pharma, Cellgene, Fresenius, Bayer Vital, Bristol-Myers Squibb, GalaxoSmithKline, Lilly, Novartis, Pfizer and Sanofi-Adventis. The author Arning was a former employee of Eli Lilly & Co and owns Lilly stock. The study was partially funded by a grant from the pharmaceutical company Lilly in Germany.

6. Assessment of the risk of bias of the study according to the RoB 2 tool

We additionally analyzed the risk of bias of the long-term study [10] using the Cochrane RoB 2 Tool [12]. Table 2 shows the risk assessment of bias according to the RoB 2 Tool [12]. Analyzing the risk of bias regarding the endpoint “overall survival” the assessment of the five bias domains led to the following results: Risk of bias was judged to be “low risk” for two of the five domains (bias due to missing outcome data and bias in measurement of the outcome) and “some concerns” for three domains (bias arising from the randomization process, bias due to deviations from intended interventions and bias in selection of the reported result). Consequently, the overall risk of bias for overall survival was “some concerns” as the study is judged to raise some concerns in at least one domain. Regarding the shortcomings of the study according to the three risk of bias domains that were judged as “some concerns” the authors came to the conclusion that these limitations are not able to alter the effects of the results of the endpoint overall survival substantially. Therefore, the risk of bias was not classified as “high”.

Table 2
Assessment of the Risk of bias of the study [10] according to the RoB 2 tool [12]

Bias domain and signalling question Response

Bias arising from the randomisation process

1.1 Was the allocation sequence random? Yes

1.2 Was the allocation sequence concealed until participants were enrolled and assigned to interventions? No information

1.3 Did baseline differences between intervention groups suggest a problem with the randomization process? No

Risk-of-bias judgement Some concerns

Bias due to deviations from intended interventions

2.1 Were participants aware of their assigned intervention during the trial? Yes

2.2 Were carers and people delivering the interventions aware of participants’ assigned intervention during the trial? Yes

2.3 If Y/PY/NI to 2.1 or 2.2: Were there deviations from the intended intervention that arose because of the trial context? No information

2.4 If Y/PY/NI to 2.3: Were these deviations likely to have affected the outcome? Probably no

2.5 If Y/PY to 2.4: Were these deviations from intended intervention balanced between groups? Not applicable

2.6 Was an appropriate analysis used to estimate the effect of assignment to intervention? Yes

2.7 If N/PN/NI to 2.6: Was there potential for a substantial impact (on the result) of the failure to analyse participants in the group to which they were randomised? Not applicable

Risk-of-bias judgement Some concerns

Bias due to missing outcome data

3.1 Were data for this outcome available for all, or nearly all, participants randomised? Yes

3.2 If N/PN/NI to 3.1: Is there evidence that the result was not biased by missing outcome data? Not applicable

3.3 If N/PN to 3.2: Could missingness in the outcome depend on its true value? Not applicable

3.4 If Y/PY/NI to 3.3: Is it likely that missingness in the outcome depended on its true value? Not applicable

Risk-of-bias judgement Low risk

Bias in measurement of the outcome

4.1 Was the method of measuring the outcome inappropriate? No

4.2 Could measurement or ascertainment of the outcome have differed between intervention groups? Probably no

4.3 If N/PN/NI to 4.1 and 4.2: Were outcome assessors aware of the intervention received by study participants? Yes

4.4 If Y/PY/NI to 4.3: Could assessment of the outcome have been influenced by knowledge of intervention received? Probably no

4.5 If Y/PY/NI to 4.4: Is it likely that assessment of the outcome was influenced by knowledge of intervention received? Not applicable

Risk-of-bias judgement Low risk

Bias in selection of the reported result

5.1 Were the data that produced this result analysed in accordance with a prespecified analysis plan that was finalised before unblinded outcome data were available for analysis? No information

Is the numerical result being assessed likely to have been selected, on the basis of the results, from:

5.2 ... multiple eligible outcome measurements (eg, scales, definitions, time points) within the outcome domain? Probably no

5.3 ... multiple eligible analyses of the data? Probably no

Risk-of-bias judgement Some concerns

Bias domain and signalling question	Response
Bias arising from the randomisation process
1.1 Was the allocation sequence random?	Yes
1.2 Was the allocation sequence concealed until participants were enrolled and assigned to interventions?	No information
1.3 Did baseline differences between intervention groups suggest a problem with the randomization process?	No
Risk-of-bias judgement	Some concerns
Bias due to deviations from intended interventions
2.1 Were participants aware of their assigned intervention during the trial?	Yes
2.2 Were carers and people delivering the interventions aware of participants’ assigned intervention during the trial?	Yes
2.3 If Y/PY/NI to 2.1 or 2.2: Were there deviations from the intended intervention that arose because of the trial context?	No information
2.4 If Y/PY/NI to 2.3: Were these deviations likely to have affected the outcome?	Probably no
2.5 If Y/PY to 2.4: Were these deviations from intended intervention balanced between groups?	Not applicable
2.6 Was an appropriate analysis used to estimate the effect of assignment to intervention?	Yes
2.7 If N/PN/NI to 2.6: Was there potential for a substantial impact (on the result) of the failure to analyse participants in the group to which they were randomised?	Not applicable
Risk-of-bias judgement	Some concerns
Bias due to missing outcome data
3.1 Were data for this outcome available for all, or nearly all, participants randomised?	Yes
3.2 If N/PN/NI to 3.1: Is there evidence that the result was not biased by missing outcome data?	Not applicable
3.3 If N/PN to 3.2: Could missingness in the outcome depend on its true value?	Not applicable
3.4 If Y/PY/NI to 3.3: Is it likely that missingness in the outcome depended on its true value?	Not applicable
Risk-of-bias judgement	Low risk
Bias in measurement of the outcome
4.1 Was the method of measuring the outcome inappropriate?	No
4.2 Could measurement or ascertainment of the outcome have differed between intervention groups?	Probably no
4.3 If N/PN/NI to 4.1 and 4.2: Were outcome assessors aware of the intervention received by study participants?	Yes
4.4 If Y/PY/NI to 4.3: Could assessment of the outcome have been influenced by knowledge of intervention received?	Probably no
4.5 If Y/PY/NI to 4.4: Is it likely that assessment of the outcome was influenced by knowledge of intervention received?	Not applicable
Risk-of-bias judgement	Low risk
Bias in selection of the reported result
5.1 Were the data that produced this result analysed in accordance with a prespecified analysis plan that was finalised before unblinded outcome data were available for analysis?	No information
Is the numerical result being assessed likely to have been selected, on the basis of the results, from:
5.2 ... multiple eligible outcome measurements (eg, scales, definitions, time points) within the outcome domain?	Probably no
5.3 ... multiple eligible analyses of the data?	Probably no
Risk-of-bias judgement	Some concerns

7. Shortcomings

By critically revising both studies following shortcomings were found:

The study design (patients with R1 resection were included) and the randomization modus were changed during the ongoing recruitment.

The small number of patients enrolled per year (0.7 patients) by each institution may indicate a selection bias.

The recruitment period had to be lengthened from 3 to 6.5 years, although high-volume centres participated.

The operative treatment, histological examination and oncological treatments were not standardized; there was no central audit.

Patients with other cancer types than adenocarcinoma were included in the study although it is known that for example tumors with neuroendocrine differentiation have a slightly better prognosis than adenocarcinomas [13].

In the first publication of the study, quality of life as a secondary endpoint was evaluated, in contrary to the second publication in which quality of life was not mentioned.

The sample size was calculated on the primary endpoint recurrence free survival and not on the secondary endpoint overall survival. So it remains unclear if the statistical power of the study was sufficient to evaluate the overall survival.

Gemcitabine treatment prolonged the median survival by only two months. Due to the intent-to-treat analysis, an unknown number of patients who received palliative chemotherapy with gemcitabine following recurrence were also included in the observation group.

38 patients were still alive in 2012: 23 in the chemotherapy group and 15 in the observation group. This is a difference of eight patients. For this difference, 186 patients were treated with chemotherapy, which could even have severe side effects.

Throughout a longer follow-up (259 events in the first analysis versus 316 events in the long-term study) the results achieve significance.

Kaplan–Meier curves differ in the two studies regarding the number of patients at risk.

Using the Cochrane RoB 2 Tool the study [10] is judged to be “some concerns” due to three of five bias domains.

Due to the long list of conflicts of interest and the partial funding by Lilly Germany, the study results were probably dependent on marketing by the pharmaceutical industry.

8. Discussion

Due to the poor prognosis of pancreatic cancer, many efforts have been made to improve disease-free and overall survival in patients with resectable adenocarcinoma of the pancreas. Radical operative techniques, perioperative chemotherapy or radio-chemotherapy, improved perioperative intensive care and improved diagnostic tools have been implemented during the last decades.

One of the treatment modules is adjuvant chemotherapy after resection of pancreatic carcinoma with curative intention. For example, the German guideline [4] recommends adjuvant chemotherapy after R0-resection in the UICC- stages I to III. On the basis of the ESPAC 3 study [1] administration of Gemcitabine is recommended with a Grad A recommendation because Gemcitabine has the same efficacy as 5-FU/folinic acid (Mayo schema) but less adverse effects. Chemotherapy should start within 6 weeks after surgery and should last 6 months. Gemcitabine is also recommended in case of R1-resection status (Grade B recommendation). Administration of FOLFIRINOX is recommended in the metastatic situation for patients with a good performance status or in a neoadjuvant concept for primary unresectable tumors of patients included in clinical trials. These recommendations correspond with the ESMO-guideline [5], the NCCN-guideline [6], etc. All of these guideline recommendations are based on a few randomized studies. We analyzed the CONKO-001 study and formerly the ESCPAC-1 study to test their validity.

As we had already discovered in the analysis of the ESPAC-1 study, the CONKO-001 study also had considerable difficulties recruiting patients. The recruitment period had to be lengthened; out of a total of 88 participating clinics over a period of 6.5 years, only 0.7 patients were randomized annually. This fact is surprising because a majority of the participating clinics were classified as high volume centers. Why was recruitment so difficult? The study did not mention whether randomization occurred before or after the surgical intervention – presumably it was after the operation, since the entire stratification refers to postoperative characteristics. Following this concept, preoperative histological confirmation of pancreatic carcinoma was not necessary, which could have made recruitment difficult.

Although the CONKO-001 study is a RCT, a relevant placebo effect bias cannot be excluded due to a lack of blinding. Manzini et al. described this effect in their analysis about the validity of studies suggesting postsurgical chemotherapy for resectable gastric cancer in 2017 [14]. Without performing a placebo control (for example throughout administrating NaCl intravenous), it is difficult to differentiate between the effect of the chemotherapy and placebo effects. Even the information regarding treatment or no treatment alone is sufficient to cause a placebo effect [15]. Often patients assigned to the observation group feel disadvantaged because they don’t receive a potential effective therapy. Without concealment of treatment allocation, the randomization procedure remains compromised because of conscious or subconscious bias [16]. This risk of bias was assessed using the Cochrane RoB 2 Tool which confirmed a risk judged as “some concerns [12]”.

The change of the randomization process during recruitment could have led to an imbalance of the two treatment groups. Anyway a simple randomization process with envelopes is not the optimal method to randomize as Jeehyoung et al. [17] and Chi-Yeon et al. [18] confirmed in their studies. The use of envelopes is not safe, because it could be possible to read the content by keeping the envelope near a light source and in this case the randomization group could be discovered and eventually changed. This affects the allocation concealment and consequently the quality of the study.

As the sample size was calculated on the primary endpoint recurrence free survival and not on the secondary endpoint overall survival, it is unclear if the statistical power is sufficient to evaluate the overall survival. For this reason, the analysis of statistical significance is possible for the primary endpoint, for the secondary endpoints descriptive statistics are indicated. Another statistical problem is the different number of patients at risk in the Kaplan Meier-curves of the primary and the long-term study.

The difference in the groups’ median overall survival amounted to around two months (20.2 months in the control group vs. 22.8 months in the therapy group). This result was significant. The difference between statistical significance and clinical relevance was discussed by Ranganathan et al., who report about common pitfalls in statistical analysis and point out that statistical significance is often misinterpreted as a clinically important result [19]. While statistical significance can be measured by p-values, clinical relevance can only be evaluated by clinicians and patients dependent on the reported effect size. Statistical significance is dependent on the study’s sample size: With large sample sizes even small treatment effects can appear statistically significant. Page [20] suggested considering measures such as effect size, confidence intervals and magnitude-based inferences instead of only rely on absence or presence of statistically significant differences. Bhardwaj et al. advises not to focus on small p-values alone but on the magnitude of treatment difference and power of the study [21].

Besides these theoretical considerations each patient will judge a prolongation of survival time of about two months differently. Quality of life and potential adverse effects under treatment have another grade of importance for patients who fight for every prolongation of survival time until death compared with patients who perhaps suffer from other comorbidities and are tired of a wearing treatment [22,23].

It is not mentioned in the study why quality of life as a secondary endpoint was not evaluated any more in the long-term study although it has been defined as a secondary endpoint in the first publication and seems to be an important factor for reasonable decision-making.

Due to the intent-to-treat analysis, an unknown number of patients who received palliative chemotherapy with gemcitabine following recurrence were also included in the observation group. It remains unknown however, how many patients of the observation group received gemcitabine for palliation and if therefore the treatment groups really differ so much.

Among the 186 patients recruited into the gemcitabine-therapy group, only six patients could not turn up for chemotherapy within the six-weeks-interval due to postoperative complications, a rate of only 3.2% which appears to be extremely low. In the ESPAC 1 study [2] 122 of 147 patients randomly assigned to receive chemotherapy were available (83 %). From these 122 patients 21 (17 %) did not receive any chemotherapy.

As there was no central audit and neither a standardized operation technique nor standardized histologic examination, the situation as described by the authors may well reflect the reality in German hospitals. However, the basic parameters should be the same for all patients when a study is conducted to test a certain therapy.

A Japanese working group examined the effect of adjuvant chemotherapy with gemcitabine in a study with a design quite similar to that of the CONKO-001 study. In 2009, the study by Ueno et al. [24] recruited 118 patients and demonstrated a significant advantage in disease-free survival in the chemotherapy group. The results for the overall survival were very similar to the CONKO-study: Median overall survival was 22.3 months in the gemcitabine group versus 18.4 in the surgery-only group, but there was a lack of significance presumably due to the smaller group of patients enrolled.

In diverse meta-analyses [25–28] which compared postoperative Gemcitabine therapy with other therapy combinations, all authors came to the conclusion that the effects of chemotherapies are altogether poor. They recommend developing further strategies besides chemotherapy to improve outcomes in the treatment of pancreatic cancer. Future studies should consider surgical interventions and a combination chemotherapy, as well as provide individualized treatment strategies based on prior chemotherapy to address the gap in knowledge of appropriate second-line therapies.

Numerous conflicts of interest are listed in the CONKO-001 study. Three authors had close contacts to pharmaceutical industries whose prosperity depends on successful approval of their products. In addition, the study was partially financed by the German manufacturer of gemcitabine. Even if the authors emphasize that the study design was not influenced by the pharmaceutical industry, the independence of the study results remains questionable.

9. Conclusion

Based on our review, the results of these two RCTs should be revisited and critically reviewed. The recommendation to include adjuvant chemotherapy with gemcitabine deserves a critical appraisal.

Footnotes

Author contributions

Bauer and Manzini contributed equally to this work; Bauer, Henne-Bruns and Manzini designed the research; Bauer and Manzini performed the research; Bauer, Henne-Bruns and Manzini analyzed the data; Manzini worked out the statistical analysis; Bauer and Manzini drafted the manuscript; Bauer, Henne-Bruns and Manzini are responsible for the interpretation of results and the critical revision of the manuscript. All authors are responsible for correctness of the statements provided in the manuscript.

Availability of data

The only data used for this study was the CONKO-001-trial [9,].

Conflict of interest

None to report.

Ethical approval

As this study is an analysis of a just existing trial this article does not contain any studies with human or animal participants performed by any of the authors.

Funding

None to report.

References

Neoptolemos

Stocken

Bassi

Ghaneh

Cunningham

Goldstein

European study group for pancreatic cancer. Adjuvant chemotherapy with fluorouracil plus folinic acid vs gemcitabine following pancreatic cancer resection: A randomized controlled trial. JAMA. 2010;304(10):1073–81. doi:10.1001/jama.2010.1275. PMID: 22782416.

Neoptolemos

Stocken

Friess

Bassi

Dunn

Hickey

European Study Group for Pancreatic Cancer. A randomized trial of chemoradiotherapy and chemotherapy after resection of pancreatic cancer. New England Journal of Medicine. 2004;350:1200–10. doi:10.1056/NEJMoa032295. PMID: 15028824.

Bauer

Weiss

Henne-Bruns

. Adjuvant chemotherapy for pancreatic cancer – can it really be recommended? Conclusions from the reassessment of ESPAC-1 study. JOP. 2018;19(1):29–35.

Seufferlein

Porzner

Becker

Budach

Ceyhan

Esposito

Guidelines Programme Oncology AWMF; German Cancer Society eV; German Cancer Aid. S3-guideline exocrine pancreatic cancer. Z Gastroenterol. 2013;51(12):1395–440. doi:10.1055/s-0033-1356220. Epub 2013 Dec 11. German. PMID:24338757.

https://annonc.oxfordjournals.org/content/26/suppl_5/v56.full.pdf+html, Cancer of the pancreas: ESMO Clinical Practice.

http://www.nccn.org/professionals/physician_gls/pdf/pancreatic.pdf.

Vera

Dotor

Feliu

González

Laquente

Macarulla

SEOM Clinical Guideline for the treatment of pancreatic cancer. Clin Transl Oncol. 2016;18(12):1172–78, Published online 2016 Nov 28. doi:10.1007/s12094-016-1586-x, PMCID: PMC5138250, PMID: 27896637.

Singapore Cancer Network (SCAN) Pancreatic Cancer Workgroup. Ann Acad Med Singapore. 2015;44(10):388–96.

Oettle

Post

Neuhaus

Gellert

Langrehr

Ridwelski

Adjuvant chemotherapy with Gemcitabine in patients undergoing curative-intent resection of pancreatic cancer. A randomized controlled trial. JAMA. 2007;297(3):267–77. doi:10.1001/jama.297.3.267. PMID: 17227978.

10.

Oettle

Neuhaus

Hochaus

Hartmann

Gellert

Ridwelski

Adjuvant chemotherapy with gemcitabine and long-term outcomes among patients with resected pancreatic cancer: The CONKO-001 randomized trial. JAMA. 2013;310(14):1473–81. doi:10.1001/jama.2013.279201. PMID: 24104372.

11.

Higgins

JPT

Altman

Götzsche

Jüni

Moher

Oxman

Cochrane bias methods group, cochrane statistical methods group. The cochrane collaboration’s tool for assessing risk of bias in randomised trials. BMJ. 2011;343:d5928. doi:10.1136/bmj.d5928.

12.

Sterne

JAC

Savović

Page

Elbers

Blencowe

Boutron

RoB 2: A revised tool for assessing risk of bias in randomised trials. BMJ. 2019;366:l4898. doi:10.1136/bmj.l4898. PMID: 31462531.

13.

Singh

Asa

Dey

Kennecke

Laidley

Law

Diagnosis and management of gastrointestinal neuroendocrine tumors: An evidence-based Canadian consensus. Cancer Treat Rev. 2016;47:32–45. doi:10.1016/j.ctrv.2016.05.003.

14.

Manzini

Henne-Bruns

Kremer

. Validity of studies suggesting postsurgical chemotherapy for resectable gastric cancer: Critical appraisal of randomised trials. BMJ Open Gastroenterol. 2017;4(1):e000138, PMID: 29177062. doi:10.1136/bmjgast-2017-000138. eCollection 2017.

15.

Waber

Shiv

Carmon

Ariely

. Commercial features of placebo and therapeutic efficacy. JAMA. 2008;299:1016–7. doi:10.1001/jama.299.9.1016, PMID: 18319411.

16.

Altman

Schulz

. Statistics notes: Concealing treatment allocation in randomized trials. BMJ. 2001;323:446–7.

17.

Jeehyoung

Wonshik

. How to do random allocation (Randomization). Clin Orthop Surg. 2014;6(1):103–9, Published online 2014 Feb 14. doi:10.4055/cios.2014.6.1.103, PMCID: PMC3942596, PMID: 24605197.

18.

Chi-Yeon

Junyong

. Randomization in clinical studies. Korean J Anesthesiol. 2019;72(3):221–32, Published online. doi:10.4097/kja.19049, PMCID: PMC6547231, PMID: 30929415.

19.

Ranganathan

Pramesh

Buyse

. Common pitfalls in statistical analysis: Clinical versus statistical significance. Perspect Clin Res. 2015;6(3):169–70. doi:10.4103/2229-3485.159943. PMCID: PMC4504060. PMID: 26229754.

20.

Page

. Beyond statistical significance: Clinical interpretation of rehabilitation research literature. Int J Sports Phys Ther. 2014;9(5):726–36, PMCID: PMC4197528. PMID: 25328834.

21.

Bhardwaj

Camacho

Derrow

Fleischer Jr.

Feldman

. Statistical significance and clinical relevance: The importance of power in clinical trials in dermatology. Arch Dermatol. 2004;140(12):1520–3.

22.

PDQ Supportive and Palliative Care Editorial Board. Planning the Transition to End-of-Life Care in Advanced Cancer (PDQ^®): Health Professional Version. 2021 Jun 11. In: PDQ Cancer Information Summaries [Internet]. Bethesda (MD): National Cancer Institute (US); 2002. Available from: https://www.ncbi.nlm.nih.gov/books/NBK223164/.

23.

Shrestha

Martin

Burton

Walters

Collins

Wyld

. Quality of life versus length of life considerations in cancer patients: A systematic literature review. Psychooncology. 2019;28(7):1367–80. doi:10.1002/pon.5054.

24.

Ueno

Kosuge

Matsuyama

Yamamoto

Nakao

Egawa

A randomised phase III trial comparing gemcitabine with surgery-only in patients with resected pancreatic cancer: Japanese Study Group of Adjuvant Therapy for Pancreatic Cancer. Br J Cancer. 2009;101(6):908–15. doi:10.1038/sj.bjc.6605256.

25.

Jin

Fan

Pan

Jin

. Gemcitabine-based combination therapy compared with gemcitabine alone for advanced pancreatic cancer: A meta-analysis of nine randomized controlled trials. Hepatobiliary Pancreat Dis Int. 2017;16(3):236–44.

26.

Jin

Teng

. Combination therapy versus gemcitabine monotherapy in the treatment of elderly pancreatic cancer: A meta-analysis of randomized controlled trials. Drug Des Devel Ther. 2018;12:475–80. doi:10.2147/DDDT.S156766. eCollection 2018.

27.

Hajatdoost

Sedaghat

Walker

Thomas

Kosari

. Chemotherapy in pancreatic cancer: A systematic review. Medicina (Kaunas). 2018;54(3):48, Published online 2018 Jul 11. doi:10.3390/medicina54030048.

28.

Matsuno

Egawa

Fukuyama

Motoi

Sunamura

Isaji

Pancreatic cancer registry in Japan: 20 years of experience. Pancreas. 2004;28:219–3.