Beyond SANO: the future of active surveillance for oesophageal cancer

The morbidity of oesophagectomy and its impact on patient’s quality of life stimulated the search for organ preservation strategies for oesophageal cancer. The Surgery As Needed for Oesophageal cancer (SANO) trial is a multicentre, stepped-wedge, non-inferiority cluster-randomized trial comparing active surveillance (AS) with standard surgery (SUR) in patients with a clinical complete response (cCR) after neoadjuvant chemoradiotherapy¹. AS was non-inferior to SUR in terms of overall survival at 2-year follow-up. By randomizing hospitals instead of individual patients and by sequential crossover of centres to the active surveillance strategy, two common problems in surgical trials were addressed. These are slow patient recruitment—a challenge that has led to premature termination of several studies comparing surgical with non-surgical treatments—and the ‘learning curve’ effect. However, the stepped-wedge design also has limitations. Clusters switch over different periods, so outcomes may be affected by calendar time rather than the intervention itself. Moreover, in the SANO trial, 35 patients from its preceding diagnostic study (preSINO) were included in the standard surgery arm. Although sensitivity analyses affirmed robustness of the outcomes, this introduced additional heterogeneity between the study arms. To our opinion, the importance of the SANO trial does not lie in providing a novel treatment standard, but in its successful execution of a crucial proof-of-concept for patients with a cCR after neoadjuvant treatment. With novel diagnostic modalities and more effective neoadjuvant treatments, the principle of active surveillance can be further investigated.

In the SANO trial, patients with adenocarcinoma (AC) were overrepresented (75%).¹ Although the results suggest that AS is equally effective for AC and squamous cell carcinoma (SCC), the trial lacked statistical power to draw definite conclusions for patients with SCC. Ongoing phase III PALACE-3 trial² will provide further clarity on the benefit and harm of active surveillance for patients with SCC.

In the AS arm of the SANO study, 48% of patients with a cCR developed a locoregional recurrence. This underlines the limitation of current response evaluations. The population of patients with cCR in the SANO trial was heterogeneous, comprising both true disease-negative patients as well as patients with minimal residual disease (MRD). MRD refers to undetectable cancer cells after treatment. These cells are seen as the root cause of tumour recurrence. Consistent with this, postoperative ctDNA positivity in esophageal squamous cell carcinoma (ESCC) has been associated with a significantly higher recurrence rate and poorer disease-free survival (DFS) and overall survival.³ Hence, delaying or omitting surgery may increase the risk of distant dissemination. Consequently, the (statistically non-significant) higher proportion of patients with distant metastases in the AS group may be a consequence of inaccurate diagnostic modalities used to select patients for AS. This concept is supported by the NEORES II trial, which questioned the safety of prolonging the interval to surgery after neoadjuvant chemoradiotherapy (nCRT).⁴

To improve organ preservation strategies, one might move beyond conventional imaging and pathological assessments for response evaluation. Liquid biopsies may play a role in assessing residual disease after nCRT. One study reported that post-nCRT ctDNA testing had a specificity of 100% and a positive predictive value (PPV) of 100% for predicting major residual disease.⁵ This implies that the detection of ctDNA almost certainly confirms the presence of residual tumour and such patients may be unsuitable for AS. However, sensitivity was only 21%, meaning that a negative ctDNA result cannot rule out residual disease. Tumour-informed personalized assays have been shown to outperform fixed-panel assays, achieving 99.2% positive detection rate at baseline and detecting ctDNA fractions as low as 0.004%. Integrating ctDNA assays with conventional diagnostics improves the sensitivity for detecting non-pathological complete response.⁶ In the preSINO trial, 28% of patients who remained ctDNA-positive after nCRT had developed distant metastases at a minimal follow-up of 12 months, in contrast to a mere 5.3% observed in the ctDNA-negative cohort.⁷ Furthermore, changes in ctDNA levels often precede radiological evidence of disease progression.³ Although it is unlikely that ctDNA can completely replace traditional biopsies and imaging, it could serve as a risk stratification tool to de-select high-risk individuals with MRD from patients who meet conventional cCR criteria. In the preSINO trial, addition of ctDNA analysis to bite-on-bite biopsies, endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA) reduced the false-negative rate from 14.9% to 5.4%.⁸ Thus, ctDNA might ‘purify’ the patient cohort entering an active surveillance protocol, thereby enhancing the safety of the strategy.

Numerous studies have shown that predictive models integrating radiomics with clinical data, using artificial intelligence and machine learning algorithms, can accurately predict pathological complete response (pCR) after nCRT. A study from China utilized deep learning-based radiomics to analyse pre- and post-treatment CT images of SCC patients. The predictive model, which also incorporated haematological indicators, had an area under the receiver operating characteristic curve (AUC) of 0.875 in the internal validation set and 0.857 in the external validation set.⁹ A combined decision-making model that uses ‘ctDNA for exclusion and radiomics for selection’ might facilitate selection of patients for AS.

Neoadjuvant immunochemotherapy and nCRT have the potential to increase pCR rates. This makes the question of organ preservation more critical than ever, but also creates a ‘paradox of success’—a dilemma for which the SANO trial offers a solution with its framework. In the era of the CROSS trial, approximately 7 of 10 patients had residual tumours, making surgery the logical default option. As pCR rates leap from 29%¹⁰ to up to 60%¹¹ nowadays, the traditional strategy of performing an oesophagectomy on all patients becomes even more questionable.

In summary, the SANO trial is a pioneering study that provided an indispensable proof-of-concept for the active surveillance strategy, opening the door for organ preservation in oesophageal cancer. However, its reliance on an nCRT regimen, the imprecise response evaluations, and the potential risks observed in the trial (such as the numerical trend towards inferior DFS) limit its direct instructional value in contemporary clinical practice.

The future direction should not be to rigidly adhere to or completely abandon SANO, but rather to build upon its foundation by constructing a new ‘SANO 2.0’ framework. The core principle of this framework is ‘precision organ preservation’: leveraging the most effective therapeutic regimens to maximize cCR rates and using the most sensitive diagnostic tools to accurately identify true complete responders. We can move beyond the past binary choice of ‘surveillance or surgery’ for an ill-defined ‘cCR’ population. It will be a truly personalized treatment pathway: one that maximizes the opportunity for organ preservation for true complete responders while ensuring that patients with locoregional residual disease receive timely surgery.

Funding

This work was supported by Applied Project of Advanced Medical Equipment.

Data availability

No data sets were generated or analysed during the current study. Therefore, data sharing is not applicable to this comment.