Multicentre cohort study to define and validate pathological assessment of response to neoadjuvant therapy in oesophagogastric adenocarcinoma

Abstract Background This multicentre cohort study sought to define a robust pathological indicator of clinically meaningful response to neoadjuvant chemotherapy in oesophageal adenocarcinoma. Methods A questionnaire was distributed to 11 UK upper gastrointestinal cancer centres to determine the use of assessment of response to neoadjuvant chemotherapy. Records of consecutive patients undergoing oesophagogastric resection at seven centres between January 2000 and December 2013 were reviewed. Pathological response to neoadjuvant chemotherapy was assessed using the Mandard Tumour Regression Grade (TRG) and lymph node downstaging. Results TRG (8 of 11 centres) was the most widely used system to assess response to neoadjuvant chemotherapy, but there was discordance on how it was used in practice. Of 1392 patients, 1293 had TRG assessment; data were available for clinical and pathological nodal status (cN and pN) in 981 patients, and TRG, cN and pN in 885. There was a significant difference in survival between responders (TRG 1–2; median overall survival (OS) not reached) and non-responders (TRG 3–5; median OS 2·22 (95 per cent c.i. 1·94 to 2·51) years; P < 0·001); the hazard ratio was 2·46 (95 per cent c.i. 1·22 to 4·95; P = 0·012). Among local non-responders, the presence of lymph node downstaging was associated with significantly improved OS compared with that of patients without lymph node downstaging (median OS not reached versus 1·92 (1·68 to 2·16) years; P < 0·001). Conclusion A clinically meaningful local response to neoadjuvant chemotherapy was restricted to the small minority of patients (14·8 per cent) with TRG 1–2. Among local non-responders, a subset of patients (21·3 per cent) derived benefit from neoadjuvant chemotherapy by lymph node downstaging and their survival mirrored that of local responders.


Introduction
Neoadjuvant chemotherapy (NAC) followed by surgery, along with perioperative chemotherapy and neoadjuvant chemoradiotherapy, is a standard of care in the management of patients with locally advanced adenocarcinoma of the oesophagus and oesophagogastric junction (OGJ) in the UK 1 . The potential benefits of NAC include: downstaging of the primary tumour 2 and lymph nodes 3 , increased tumour resectability 4 , elimination of micrometastases 5 and improved survival 6 . Early assessment of response to NAC may provide information to tailor multimodal therapy 7 .
Both NAC and surgery are associated with considerable morbidity and mortality 8 , and evidence remains inconsistent for the survival benefit for patients who undergo NAC 4,8,9 . The most recent meta-analysis 6 to compare NAC versus surgery alone in 2062 patients suggested a 5⋅1 per cent absolute survival advantage at 2 years for patients treated with NAC for adenocarcinoma. This is because only a small minority of patients have a significant pathological response to neoadjuvant therapy and it is these patients who gain a significant survival benefit from NAC 10,13 .
There are numerous methods for assessing pathological response to neoadjuvant therapy, but no universal measure is used consistently 10,14 -18 . The majority were developed for patients who underwent neoadjuvant chemoradiotherapy and did not differentiate patients based on histology. Few have been validated in patients with oesophageal adenocarcinoma undergoing NAC 2,19 -22 . Use of the Tumour Regression Grade (TRG) described by Mandard and colleagues 23 is suggested by UK guidelines, although this has not gained universal acceptance, and the guidelines give no detail regarding how TRG should be used to guide therapy decisions 1,17 . This system is based on the amount of residual tumour and degree of fibrosis at the primary tumour site, which is graded on a five-point scale 23 . Reports from single-centre cohorts 20,24 and from small subsets of larger multicentre trials 25 have identified a significant survival advantage with Mandard TRG 1-2 or TRG 1-3, further confusing clinical decision-making.
A number of clinically important questions could be addressed by a robust and universally accepted measure of response to neoadjuvant treatment, including the development of biomarkers to accurately predict an individual patient's tumour response to preoperative therapy; this would lead to non-responders proceeding directly to surgery or being considered for alternative neoadjuvant regimens, and the identification of patients who are likely to benefit from adjuvant therapy in new stratified trials.
This multicentre cohort study evaluated the current status of neoadjuvant response assessment in multidisciplinary team decision-making via a questionnaire, and aimed to define and validate the pathological assessment of response to NAC in treated oesophagogastric adenocarcinoma. The aim was to provide a consistent, simple, robust and universally acceptable method to assess response to neoadjuvant therapy to allow wider application for both clinical use and biomarker discovery.

Questionnaire
A questionnaire was distributed to 11 upper gastrointestinal cancer centres, all part of the Oesophageal Cancer Clinical and Molecular Stratification (OCCAMS) consortium, to assess the current use of neoadjuvant response assessment in clinical practice (Appendix S1, supporting information). The OCCAMS consortium is a UK-wide multicentre consortium to facilitate clinical and molecular stratification of oesophagogastric cancer, with ethical approval for biological sample collection and analysis in conjunction with detailed clinical annotation (Research Ethics Committee number 10/H0305/1).

Patients
The records of consecutive patients undergoing oesophagogastric resection (tumours of the oesophagus and OGJ only were included) treated at seven centres (University Hospital Southampton NHS Foundation Trust, Belfast Health and Social Care Trust, University Hospitals Birmingham NHS Foundation Trust, University Hospital Cambridge NHS Foundation Trust, Royal Infirmary of Edinburgh, Portsmouth NHS Trust, Nottingham University Hospitals NHS Trust) between January 2000 and December 2013 were reviewed as part of the OCCAMS consortium. All patients were discussed at a specialist multidisciplinary team meeting. Standard staging investigations included high-resolution CT, endoscopic ultrasonography, and latterly integrated fluorodeoxyglucose PET-CT and staging laparoscopy, where indicated. Patients deemed suitable for potential surgical resection with tumours staged as cT2 Nx M0 or cTx N+ M0 were considered for NAC based on local practice and national guidelines 1 .
CT or PET-CT was repeated before surgery to assess the response to chemotherapy and disease operability.
Data recorded included demographics, tumour characteristics, resection type and histopathological analysis of the surgical specimen. The TNM classification (7th edition) was used to report tumour stage after analysis of pathology reports 26 . Pathological tumour clearance (R status) was determined according to the Royal College of Pathologists' guidance.
Overall survival (OS) was defined as time from operation to date of death from any cause or date of last review.

Factors analysed
Pathological response to chemotherapy was assessed using the TRG system, with regression based on the degree of fibrosis and residual cancer cells (TRG 1-5) 23,27 . TRG was scored by specialist gastrointestinal pathologists blinded to the clinical data at the treating cancer centre. Some 10 per cent of cases were validated externally by an independent pathologist as part of the OCCAMS/International Cancer Genome Consortium project 28,29 with a κ value exceeding 0⋅8. All dissected lymph nodes were stained with haematoxylin and eosin, and analysed microscopically for metastatic disease. Lymph node downstaging was defined as any regional lymph node that was positive on clinical evaluation (cN+) which subsequently had no evidence of pathological regional lymph node disease (ypN0), as described previously 24 .

Statistical analysis
Descriptive data are presented as median (range) unless stated otherwise. Data were analysed using the Kruskal-Wallis, Mann-Whitney U and Pearson's χ 2 tests, as appropriate. Kaplan-Meier analysis, and univariable and multivariable Cox logistic regression modelling were used to assess the relationship between pathological response grading systems and OS. All factors that showed statistical significance in univariable analysis were entered to derive the final model. Stratified analyses were performed based on receipt of NAC, nodal stage and response to chemotherapy. P < 0⋅050 was considered statistically significant for all tests. Statistical analysis was undertaken in SPSS ® version 22 (IBM, Armonk, New York, USA).

Current clinical use of response assessment to neoadjuvant chemotherapy
The responses from 11 UK cancer centres demonstrated that TRG was the system most widely used to assess response to NAC (used in 8 of 11 centres), and felt to be useful in providing prognostic information for the patient (8 of 11) and making decisions about the modification of adjuvant therapy (9 of 11). There was no consensus on how TRG was being used to influence decision-making for individual patients in practice; the centres used different scores to define responders, with most using TRG 1-3 (5 of 8), and there was a lack of consensus on how adjuvant therapy should be guided by TRG. Some centres would advocate adjuvant therapy based on whether the patient had responded to therapy (5 of 11), whereas others would not use response information (6 of 11) (Fig. S1, supporting information).    of Pathologists was achieved in 66⋅6 per cent (913 of 1371) and the median nodal yield was 23 (0-75). Patient characteristics, and clinical and pathological outcomes are summarized in Table 1.
Responders and non-responders had similar preoperative clinical features (age, sex) and clinical stage of disease, yet responders had markedly reduced ypT category (P < 0⋅001), ypN status (P < 0⋅001) and were more likely to have nodal downstaging (P < 0⋅001) ( Table 3). Complete resection (R0) was achieved in 92⋅5 per cent of responders (173 of 187) compared with 62⋅5 per cent of non-responders (678 of 1085), and this correlated across the five TRG groups (Table S1, supporting information). Of patients who underwent an R1 resection, tumour involvement was at the radial (circumferential) margin in 92⋅7 per cent, regardless of treatment centre or type of surgery performed. There was no significant difference in nodal yield between responders and non-responders (P = 0⋅437).

Evaluation of chemotherapy regimen
Patients treated with platinum-based triplet chemotherapy had a significantly greater response to chemotherapy in the local tumour (TRG) and regional lymph nodes, and were more likely to have an R0 surgical resection than those who received cisplatin and 5-FU (Table 3; Table S2, supporting information).
Kaplan-Meier analysis showed better OS among patients who had platinum-based triplet chemotherapy (Fig. S2, supporting information), but multivariable analysis showed no difference in OS between regimens ( Table 4).
Values in parentheses are 95 per cent confidence intervals. TRG, Tumour Regression Grade.

Discussion
In this multicentre study, a clinically meaningful local response to NAC for adenocarcinomas of the oesophagus and OGJ was restricted to the small minority of patients (14⋅8 per cent) with TRG 1-2. Among apparent local non-responders, there was a subset of patients who appeared to derive additional benefit from NAC by lymph node downstaging and their survival mirrored that of local responders. The difficulties faced by clinicians in routine practice regarding what constitutes a meaningful response to NAC, and how this information should be used to tailor subsequent treatment, has been caused primarily by the cohort sizes of previous studies. For example, the Mandard system was developed in 1994 in a cohort of 93 French patients (84 per cent with squamous cell cancer) treated with cisplatin and radiotherapy, and TRG 1-3 was found to correlate with improved disease-free survival 23 . In a subsequent study performed in the UK and Ireland 30 , TRG had no correlation with survival in 43 patients with adenocarcinoma. More recent work from single institutions has demonstrated the validity of the TRG system in patients with oesophageal cancer treated with NAC, but cohort sizes remain small (Fareed et al. 19 : 103 patients, TRG 1-3 associated with a disease-specific survival advantage; Noble et al. 24 : 136 patients, TRG 1-2 associated with a disease-free survival advantage). Larger series have focused on the role of NAC and tumour stage rather than TRG 31 , or have included mainly gastric cancers 25 . In the context of neoadjuvant chemoradiotherapy a three-point scale for TRG, using TRG 1 compared with TRG 2-3 and TRG 4-5, was found to provide the best discriminant fit of all response measurement modalities in a cohort of 393 patients from a single centre in Ireland 32 . The data presented here do not support this classification, and suggest that there may be differences in the histomorphological assessment of response between chemotherapy and chemoradiation.
The strengths of the present study are its cohort size, length of follow-up and multicentre nature. It shows that TRG is a robust measure of local response to NAC in routine clinical practice, with excellent correlation between local centre scoring and central validation. The inclusion of the two UK centres (Nottingham 19 and Southampton 24 ) that previously published discordant results regarding the level of TRG associated with a response adds weight to the finding that only TRG 1-2 represents a group of true local responders. This is supported by the use of OS as the outcome measure here, and by the similarity between responder and non-responder groups in terms of pretreatment characteristics.
In contrast to the results of a recent subgroup analysis of the Medical Research Council (MRC) Adjuvant Gastric Infusional Chemotherapy (MAGIC) Trial 25 , where lymph node status was the only independent predictor of survival in patients treated with chemotherapy, in the present study non-response to NAC (TRG 3-5) was independently associated with poor overall survival. This probably reflects the relative sizes of the study cohorts (1293 in the present study versus 330 in MAGIC 25 ). In the present study, no attempt was made to assess the value of postoperative chemotherapy in either responders or non-responders. In a recent study of 333 patients from a single UK centre 33 , only responders to NAC were observed to derive a survival advantage from the adjuvant part of the MAGIC regimen, and there was evidence of potential harm for non-responders in terms of chemotherapy morbidity. The question of who should receive adjuvant treatment and what form this should take needs to be addressed urgently in prospective studies, so that futile overtreatment with chemotherapy can be avoided and therapies can be targeted more effectively where appropriate. TRG and lymph node downstaging could be used to stratify patients while the validation of recently discovered mutational endotypes takes place 28 .
This study was not designed to investigate differences in outcome between different chemotherapy regimens. However, the data clearly show the superiority of platinum-based triplet chemotherapy, including epirubicin, over cisplatin and 5-FU for local tumour response, but this did not translate into better survival. These findings are in keeping with the results of the MRC OEO5 trial 34 , in which two cycles of neoadjuvant cisplatin and 5-FU was shown to be equivalent to four cycles of ECX for overall survival, with higher chemotherapy-related toxicity in the ECX group 34 . Widely regarded as a negative trial, OEO5 is important because it identifies the requirement for robust markers of patient and tumour stratification to guide precision treatment.
There are clear drawbacks when performing a large multicentre cohort study over a relatively long period. This was not a randomized trial and there are missing data, so bias cannot be excluded, but the sample size helps to negate this deficiency. Staging modalities, chemotherapy and to a lesser extent surgery will have changed over the study interval. It is possible that a number of patients who received treatment at the beginning of the study may have been excluded from treatment had they been staged using modern imaging, leading to worsened OS. It could be presumed that these patients would be in the non-responder group, as there is some evidence to support the association between local tumour response and systemic relapse, but tumour stage after NAC seems to be more important than initial stage at presentation in terms of assessing prognosis 31 . These drawbacks may also explain the relatively low overall R0 resection rate (66⋅6 per cent), but it is important to note that the circumferential margin was involved in the majority of R1 resections and the more stringent Royal College of Pathologists' definition of margin involvement was used. No attempt has been made to assess the theoretical benefit of using NAC (rather than neoadjuvant chemoradiotherapy) to treat distant micrometastatic disease in these patients. Ongoing randomized studies will hopefully answer this important question.
The finding that a small group (approximately 20 per cent) of patients whose primary tumours apparently do not respond to NAC have downstaging of local nodes and associated good long-term survival, similar to that of primary tumour responders, is important both for discussions of prognosis with individual patients and for the design of the next generation of trials of tailored adjuvant treatment. The use of cN category to assess involvement of nodes before operation compared with postoperative ypN status to determine downstaging is open to criticism. In support of this strategy, it has been shown previously that patients with ypN0 disease have worse OS than those with pN0 disease, suggesting true nodal involvement 24 , and downstaging has been demonstrated accurately in other cohorts 31 . This also reflects current clinical practice in the UK and elsewhere. However, an alternative interpretation of these data is that they are indicative of clinical overstaging rather than downstaging. A future analysis should consider the pathological assessment of nodal downstaging to look for evidence of fibrosis/previous tumour in the nodes, and the relationship between this and prognosis. Large-scale collaborations such as the OCCAMS consortium are ideally placed to do this.
As the research community begins to consider the move from binary 'one size fits all' treatment and trial designs to more personalized strategies, robust markers of treatment response will be required. The present findings confirm TRG as such a marker and clearly define groups of patients who benefit from NAC. In addition, giving clarity to the assessment of response offers the opportunity to determine biomarkers that may predict response to existing and novel neoadjuvant treatments, whether they are patient-, tumour-or treatment-related.