Tumour markers in lung cancer: modelling strategies for interpretation of tumour marker changes and potential clinical applications

By Dr F. van Delft, Dr M. Schuurbiers, Dr van den Heuvel and Dr H. van Rossum

The development of immunotherapy and other targeted therapies has given rise to significant improvements in the treatment of non-small cell lung cancer. However, some patients do not benefit from these therapies and it is, therefore, important for a variety of reasons to stop treatment. This article discusses the development of a ‘Serum Tumor Marker-based Outcome Prediction’ (STOP) test that enables identification of patients with no durable benefit from treatment as early as 6 weeks after treatment initiation.

Lung cancer

Lung cancer is the most deadly cancer worldwide, with approximately 2 million deaths annually. The two most common types of lung cancer are non-small cell lung cancer (NSCLC) and small cell lung cancer (SCLC), with the former accounting for approximately 85% of all lung cancer cases. A major reason for its lethality is that lung cancer is usually diagnosed at an advanced stage with lymph node involvement and/or metastases. Owing to the historical lack of successful systemic treatments, the survival of patients with advanced lung cancer (NSCLC) has been very limited with a 5-year survival rate of less than 10% for patients with distant metastases.

Over the past 10 to 15 years, there have been some significant improvements in the treatment of advanced NSCLC with the introduction of targeted therapies and immunotherapy. Targeted therapies act on specific oncogenic drivers that are responsible for tumour initiation and growth. Relevant oncogenic drivers for lung cancer include epidermal growth factor receptor (EGFR) and anaplastic lymphoma kinase (ALK) mutations and alterations, which can be treated with specific small molecule inhibitors. For lung cancer, immunotherapy is based on immune checkpoint inhibitors, which stimulate the patient’s immune system to attack the lung cancer, generally by interfering with the programmed cell death protein 1 (PD-1)/programmed death-ligand 1 (PD-L1) checkpoint proteins.

The availability of these new treatment options and the ongoing tremendous development of new treatment strategies has changed the therapeutic landscape for advanced lung cancer. In particular for patients who are eligible for targeted treatment, multiple lines of effective treatment have become available. Although these new treatments have revolutionized lung cancer care, these new drugs are associated with high costs and potentially serious and sometimes fatal side effects. In addition, a large group of patients, especially those receiving immunotherapy, do not show a durable therapy response and therefore unfortunately do not benefit from this treatment. All these limitations have created an interest and need for biomarkers that can predict treatment response, monitor response, and allow early
detection of recurrence after initial treatment response.

shutterstock 2229048487

Illustration of lung cancer (Shutterstock.com)

Tumour markers in lung cancer

In the field of lung cancer, the majority of biomarker research is focused on new technologies and biomarkers in blood such as circulating tumour DNA (ctDNA), commonly referred to as ‘liquid biopsy’. Biomarkers to detect ctDNA represent a highly heterogeneous group of markers including biomarkers that focus on analysing gene variations and alterations by using different methods (from single mutation sites to gene panel to whole exome or genome analysis), ctDNA methylation profiles, and fragment size analysis. Challenges associated
with ctDNA analysis include the wide variation and critical aspects of the preanalytical workup, lack of standardization and harmonization of both preanalytical, analytical and post-analytical procedures, the technical skills needed to perform these tests, and the high costs. Overall, there are still significant steps required before ctDNA can be widely adopted in clinical practice as a biomarker suitable for cancer follow-up.

Besides ctDNA, there are other tumour markers in patients with (advanced) NSCLC, including cancer antigen 125 (CA125),carcinoembryonic antigen (CEA), cytokeratin 19 fragment (CYFRA 21-1), neuron-specific enolase (NSE) and squamous cell carcinoma antigen (SCC). Some of these tumour protein markers are commonly available in clinical laboratories and are fairly well characterized in terms of antigen recognition and binding sites, analytical performance specifications and diagnostic performance for (advanced) lung cancer. Furthermore, their diagnostic value as prognostic and predictive markers has been clearly established. Therefore, this group of authors, together with many others, has investigated the use of these readily available tumour markers for the follow-up of advanced NSCLC.

A major insight that has been identified is that when using tumour markers for cancer follow-up, the change in tumour marker concentration over time is more relevant than the concentration itself, as this is most likely to reflect changes in tumour burden. To date, very limited studies have been conducted on methods and models to optimize and validate such longitudinal or sequential tumour marker models.

Schermafbeelding fig 1

Figure 1. Biomarker response characteristic plot for CEA in NSCLC
Based on a cohort of 344 advanced non-small cell lung cancer (NSCLC) patients treated with immunotherapy the CEA change between base-line and 6 weeks of follow-up is plotted against a clinical responses observed at 6 months [1,2].

Sequential measurement of tumour markers

The traditional ways of interpreting sequential tumour marker measurements are either to compare the values obtained to reference ranges, or to rely on the physician’s experience as to what the results and their changes mean clinically. Another method generally recommended in the field of laboratory medicine is the use of critical differences based on both method imprecision and intra-individual biological variation. However, for tumour markers there are some highly relevant limitations of this approach, including the lack of correlation with relevant clinical endpoints such as response or non-response to treatment, the uncertainty introduced by the biological variation of tumour markers in cancer patients, and the lack of reference values specifically for cancer patients, as these are generally based on healthy volunteers [1]. Clinical utility and impact are also not considered. Clinical utility refers to the improvement of either
the patient’s clinical performance (in oncology preferably survival), or quality of life. To overcome the issues raised and potentially improve the clinical performance of cancer patients, new approaches and methods to interpret, optimize and validate the use of sequential tumour marker measurements to predict relevant clinical events, are essential. The authors have therefore used retrospective tumour marker results from NSCLC patients receiving immunotherapy to investigate how these might reflect the clinical benefit of these patients by using new modelling strategies.

Scherm­afbeelding 2023 11 08 om 10.23.21

Figure 2. Diagnostic performance of initial training and validation of the STOP model

Adapted from van Delft et al. [5], development of the Random Forest model, which forms the basis for the STOP model. A bootstrap with 1000 iterations each containing 75% of patients was performed to assess the effects of patient selection.

Clinical application, BREC-plot and advanced modelling of sequential tumour marker data

Initially, the best achievable application of tumour markers was considered to be the prediction of non-durable benefit (NDB) in a selection of the NSCLC patients treated with immunotherapy. This was considered to allow early discontinuation of ineffective treatment (reducing side effects and costs) and timely initiation of a next, potentially more effective treatment. To enable a reliable discontinuation of immunotherapy, the accuracy of the predicted NDB had to be high and therefore the minimum test specificity of the to-be-developed test, was set at 95%. Optimization and training of tumour marker models were then primarily focused on this potential clinical application and set requirements. The clinical definition of NDB was defined as  death, progressive disease objectified by imaging, or clinical progression determined by the treating physician within 6 months after treatment initiation. As a first, rather simplified approach, the biomarker response characteristic plot (BREC)-plot was developed to graphically represent the clinical significance of relative changes in tumour marker concentrations (Fig. 1) [2]. Using this approach, some simple logical tumour mark cut-offs for CEA, CYFRA 21-1 and NSE were identified that could predict NDB in approximately 32.3 % of NSCLC patients [3].

However, the use of the rather simple rules was considered to limit the potential performance of such a test and, as a next step, more advanced modelling techniques were investigated. First, modelling strategies focused on individual tumour markers were investigated, ranging from simple logical models to machine learning and AI methods [4]. Next, primarily AI-based methods (including neural networks, boosting, and random forest methods) were used to investigate their suitability for automatically optimizing models that interpret the changes in multiple tumour markers together [5]. Finally, a model was selected, a random forest model based on CYFRA 21-1, CEA and NSE results (Fig. 2), which provided the best diagnostic performance, to be further validated in a new and independent cohort of NSCLC patients reflecting the current medical practice. In this independent cohort of 242 patients, the newly assigned ‘Serum Tumor Marker-based Outcome Prediction’ (STOP) model, enabled the detection of NDB with 38.1% sensitivity and 95.9% specificity, and was found to have a similar diagnostic performance to imaging performed at the same time (6 weeks after treatment initiation). Most importantly, only by combining imaging and the STOP test enabled identification of patients with NDB as early as 6 weeks after treatment initiation with 100% specificity and a positive predictive value of 100%; when both indicated progressive disease at 6 weeks none of the patients benefited from treatment based on the criteria used. Overall, using the latter criterion, 21% of the patients with NDB were accurately identified.

Summary and conclusions

The studies performed showed that the use of the more traditional tumour markers, such as CEA, CYFRA 21-1 and NSE, can provide relevant monitoring support for NSCLC patients. In addition, new methods were developed to allow more advanced modelling of sequential (tumour marker) data, which improved the diagnostic performance of the tumour marker models. Future research and development in this area should focus on further developing powerful mathematical methods for interpreting sequential data and to applying them to other longitudinal (tumour marker) data, such as those available for other types of cancer. For the developed STOP model, an additional blinded multicentre prospective clinical validation would be of added value to ultimately confirm the presented results and strengthen its potential place in NSCLC clinical care.

The authors

Freek van Delft1,2 PhD; Milou Schuurbiers3 MD; Michel van den Heuvel3 MD, PhD; Huub H. van Rossum1* PhD, EuSpLM

1 Department of Laboratory Medicine, The Netherlands Cancer Institute, Amsterdam, 1066CX, The Netherlands
2 Health Technology and Services Research Department, Technical Medical Centre, University of Twente, Enschede, The Netherlands
3 Department of Respiratory Diseases, Radboud University Medical Center, 6500 HB, Nijmegen, The Netherlands

* Corresponding author Email: h.v.rossum@nki.nl

1. van Rossum HH, Meng QH, Ramanathan LV, Holdenrieder S. A word of caution on using tumor biomarker reference change values
to guide medical decisions and the need for alternatives. Clin Chem Lab Med 2022;60(4):553–555 (https://www.degruyter.com/document/doi/10.1515/cclm-2021-0933/html).
2. Moritz R, Muller M, Korse CM et al. Diagnostic validation and interpretation of longitudinal circulating biomarkers using a biomarker
response characteristic plot. Clin Chim Acta 2018;487:6–14 (https://www.sciencedirect.com/science/article/abs/pii/S0009898118304923?via%3Dihub).
3. Muller M, Hoogendoorn R, Moritz RJG et al. Validation of a clinical blood-based decision aid to guide immunotherapy treatment in patients with non-small cell lung cancer. Tumour Biol 2021;43(1):115–127 (https://content.iospress.com/articles/tumor-biology/tub211504).
4. van Delft FA, Schuurbiers M, Muller M et al. Modeling strategies to analyse longitudinal biomarker data: an illustration on predicting immunotherapy non-response in non-small cell lung cancer. Heliyon 2022:e10932 (https://www.cell.com/heliyon/fulltext/S2405-8440(22)02220-4?_returnURL=https%3A%2F%2Flinkinghub.elsevier.com%2Fretrieve%2Fpii%2FS2405844022022204%
5. van Delft FA, Schuurbiers MMF, Muller M et al. Comparing modeling strategies combining changes in multiple serum tumor biomarkers for early prediction of immunotherapy non-response in non-small cell lung cancer. Tumour Biol 2023. doi: 10.3233/TUB-220022 (online ahead of print).