¹⁸FDG-PET-CT identifies histopathological non-responders after neoadjuvant chemotherapy in locally advanced gastric and cardia cancer: cohort study

Cancer of the stomach (GC) and the distal esophagus and esophagogastric junction (AEG) is a substantial global health problem with around 1 million new cases and 750,000 deaths per year, accounting for estimated 10% of all cancer-related deaths [1, 2]. In Europe and North America, the overall 5-year survival for GC is approximately 25% [3], while superior outcomes with 5-year survival rates of approximately 60% are reported in East Asia [4].

The optimal medical treatment for advanced GC and AEG is still a source of debate, but after the publication of the randomized “MAGIC Trial” and “ACCORD Trial”, neoadjuvant chemotherapy has become first choice for the treatment of locally advanced GC and AEG, and reported improved 5-year survival rates of 36 and 38% respectively [5, 6]. This situation is similar for esophageal and cardia cancers where the recently published randomized “CROSS Trial”, using neoadjuvant chemoradiation, reported 3-year survival rates of 59% [7].

Objective assessment of the treatment effect after neoadjuvant treatment is only possible by histopathology in the resected specimens. In patients with AEG and GC, the presence of < 10% of vital residual tumor cells is considered as a major pathologic response, and is associated with a significant survival benefit [8–10]. The difficulty is to identify patients who do not respond or progress under neoadjuvant treatment. Those patients may not profit from neoadjuvant treatment, still suffer from adverse events, and finally risk tumor progression.

In the setting of esophageal cancer, measurement of early changes in tumor glucose uptake by use of 18-fluorodeoxyglucose-PET (PET) and later PET-CT yielded promising results for predicting response following neoadjuvant chemotherapy [11]. Metabolic tumor activity can be quantified by the standardized uptake value (SUV), and it was shown that a drop of ≥35% measured after 2 weeks of induction chemotherapy was an accurate cut-off value to predict response [12]. This cut-off at ≥35% was prospectively studied in the MUNICON-I trial including 119 patients with AEG I and II [13]. Metabolic response evaluation by PET-CT accurately identified all non-responding tumors within two weeks of treatment. In addition, there was a significant survival difference between metabolic responders and non-responders.

No data are currently available for early metabolic response evaluation by PET-CT in patients with AEG III and GC, and the potential benefit is therefore unclear. We studied patients with AEG II/III and GC in a cohort of patients using the criteria for early metabolic response from the MUNICON-I trial to evaluate the accuracy and feasibility of metabolic response evaluation by early PET-CT following neoadjuvant CTX.

A retrospective cohort of 72 consecutive patients with biopsy proven GC or AEG Siewert type II-III [14] was included in this study. All patients underwent routine staging, including, laboratory tests, upper GI-endoscopy with endoscopic ultrasound (EUS) and ¹⁸FDG PET-CT as reported previously [15]. Patients with stage cT2N+ or cT3–4, Nx by EUS and PET-CT (UICC TNM Classification, 7th edition, [16] underwent diagnostic laparoscopy to exclude occult peritoneal carcinomatosis prior to neoCTX. All patients were discussed in a multidisciplinary specialized tumor board prior to treatment initiation. The study was approved by the local ethics committee (PB 2016–00769).

This cohort study in patients with locally advanced AEG II/III and GC shows that metabolic response two weeks after starting neoadjuvant therapy evaluation by FDG-PET-CT, using the validated threshold for metabolic responders from the MUNICON-I intervention trial (13) of ≥35% reduction of SUV does not accurately predict overall pathologic response. It does however identify a subgroup of patients that does not respond to neoCTX with a specificity of 90%. In this cohort, this subgroup compromised 14% (10/72) of the study population with an inferior prognosis compared to PET responders.

The results of our study compare well to the results of the MUNICON-1 trial in patients with esophageal cancer, where 110 patients with AEG I/II were evaluated by PET-CT [13]. The authors claim that responders can be identified by early metabolic imaging, however, 42% of the 50 PET responders showed a minor regression like the 50% observed in our study. The NPV for the metabolic response was 100% in the MUNICON trial, which is comparable to the 90% in our study and the 10% difference is likely due to the low patient numbers and different tumor entities. We therefore conclude that metabolic response evaluation by PET-CT does not accurately predict overall response but identifies non-responders in both trials. We also confirmed that PET-responders have a better prognosis than non-responders (p = 0.04) with remarkably close survival rates (median survival was not reached in PET-responders in both trials and 26 months and 28.2 months respectively for PET non-responders). In the MUNICON-I intervention trial, chemotherapy was discontinued in metabolic non-responders, thereby saving time, and reducing side-effects and costs without compromising the outcome.

Less data is available for patients with gastric cancer. Vallböhmer et al. [22] found no predictive value for the FDG uptake in 40 gastric cancer patients. Ott and colleagues [23] prospectively studied 49 GC patients including Siewert type III tumors with a metabolic response (SUV reduction ≥35%) by PET. Overall, 23/49 (47%) patients had non-intestinal type cancers and 38/49 (78%) were in the proximal third. Metabolic response correctly predicted histopathologic regression in 11/16 responding and 27/33 non-responding tumors. This resulted in a sensitivity and specificity of 69 and 82%. PPV and NPV were 65 and 84% and overall accuracy was 78%. Median survival of metabolic responders was not reached, and significantly (p = 0.037) better than for non-responders (24.1 months). Again, remarkably similar results were obtained in our study. The lower NPV is likely attributable to the higher proportion of non-intestinal tumors (46.9%) compared to our study (20%) and the significantly lower baseline SUV obtained in non-intestinal-type tumors.

In contrast to patients receiving chemotherapy alone, early metabolic response evaluation by PET-CT was not successful in patients receiving neoadjuvant chemoradiation for esophageal and esophagogastric junction cancers. [24–26]. Technically, a higher cut-off value might be better to predict histopathological response. Indeed, a previous study from the Munich group [12] showed that a 45% or more decrease in SUV would result in higher specificity for histological response (86% versus 75%) but most important, in a lower NPV. This was also true in our study with a cut-off at 50% (data not shown). The FDG uptake is not uniform among the subgroups. Esophageal tumors, AEG I, show about 100% FDG uptake, much higher than AEG III or GC, or diffuse type cancer [15]. Therefore, the use of PET-CT is not uniformly recommended, particularly in non-intestinal gastric cancer including signet ring cell cancer [23, 27–29].

Distal gastric cancer with low differentiation grades, including diffuse types are less likely to achieve major tumor regression after chemotherapy [30]. A risk score was evaluated in 410 patients receiving neoCTX for GC. Well-differentiated tumor grading, intestinal tumor type histology and tumor localization in the middle third of the stomach were identified as the significant positive predictive factors for histopathologic response and prognosis. A prognostic index could be created based on tumor localization, grading and type according to Lauren classification that identified 3 risk groups (low, intermediate and high) with significantly different clinical and histopathological response rates and overall survival times. [31]. Several molecular markers have been investigated in view of characterizing tumor entities and predicting tumor response and prognosis following neoadjuvant treatment [32, 33], since a variety of novel targeted therapeutic approaches are introduced in cancer treatment [34]. In HER2-positive advanced GC and AEG, the international phase III trastuzumab for GC (ToGA) study showed a significant improvement in the median overall survival of patients upon the addition of trastuzumab to cisplatin and fluoropyrimidine backbone therapy [35]. In the MUNICON-II trial, salvage neoadjuvant radiochemotherapy in metabolic non-responders lead to local remissions in a considerable number of patients but was not able to change the clinical course [36]. Altogether, metabolic non-responders may profit from a therapeutic switch to these novel approaches.

Our study clearly has its limitations, mainly attributable to the small study population which does not allow more sophisticated statistical analysis including multivariate testing, and all test results must therefore be interpreted with caution. Despite these limitations, our results are remarkably close to the results of two published comparable studies, the MUNICON PET-CT trial in Siewert type I/II tumors [13] and the PET-study in gastric cancer including Siewert type III tumors [23] using early metabolic response evaluation. Furher research could also include assessing survival outcomes in patients classified as non-responders by PET-CT with subsequent discontinuation of neoadjuvant chemotherapy and comparing them with those remaining on treatment.

In conclusion, our study in patients with AEG and GC adds further evidence that response PET-CT reliably detects metabolic non-responders and can therefore identify patients who should either immediately proceed to resection or receive a modified multimodality therapy. PET-CT-guided neoadjuvant chemotherapy appears feasible in patients with AEG and GC but important issues remain to be addressed in future trials especially standardization for metabolic imaging as planned by the EORTC GI Group and NEOPEC Trial Group [37, 38].