The prognosis for pancreatic cancer patients is very poor, with only around five percent living beyond five years following diagnosis. Current systemic treatments, such as chemotherapy, are limited in effect, making pancreatic cancer the fourth leading cause of cancer death. In recent years the molecular characterisation of cancers has improved our understanding of their heterogeneity and potential sensitivity to treatment. Both breast and gastric cancer have been shown to include tumours that exhibit amplification of the HER2 gene and have consequently been treated with monocloncal antibodies targeting the HER2 receptor. Andrew Biankin from Kinghorn Cancer Centre and Garvan Institute of Medical Research, Australia, Sean Grimmond from the University of Queensland, Australia and colleagues set about characterising HER2 amplification in pancreatic cancer to better understand its clinical implications, as published in a recent study in Genome Medicine. Biankin and Grimmond explain more about their findings and the challenges of using molecular data in a clinical setting.
Why are you interested in understanding pancreatic cancer and specifically its molecular profile?
Pancreatic cancer is a lethal malignancy. It is the fourth leading cause of cancer related death in western societies, and the survival rate has not changed for almost 50 years. Whilst some systemic therapies are used in the treatment of this disease, surgery remains our only real option for a cure. Unfortunately only a low proportion, about 15-20 percent, are suitable for operative resection at the time they are diagnosed. Even so, 80 percent or more still die of the disease having undergone potentially curative resection.
Although systemic therapies are largely ineffective, we do see significant responses in some patients, and in some instances exceptional responses. Due to our evolving understanding of cancer in general, and its molecular diversity and heterogeneity, even in microscopically similar cancers, we hypothesize that pancreatic cancer may not have many current effective therapies because of this diversity. Identifying these sub-groups based on their molecular characteristics for us suggests that we could better stratify patients for appropriate therapies.
What would you say are the most interesting findings of your study?
What we consider to be the most interesting and exciting findings of our study is that these types of cancers, i.e. HER2 amplified pancreatic cancers, appear to have different clinical characteristics. This has significant implications in practice because most pancreatic cancers are expected to metastasize first to the liver. In HER2 amplified pancreatic cancer there appears to be a lack of metastases to the liver and more metastases to the lung and perhaps the brain. This is important because it may be difficult to identify the primary source of disease in a setting where you have multiple lung metastases. The pancreas is often difficult to visualise, and identifying a primary cancer, which may often be relatively small, can be difficult. In addition, with regard to surveillance for recurrent disease after pancreatic resection, looking at other non-traditional sites such as the lung and the brain for metastases would lead to earlier detection of recurrent disease.
In broader terms, the lack of liver metastases in HER2 amplified cancer suggests that organ specific metastases relate to biology rather than anatomy. Traditionally it was believed that most tumours metastasized to the liver because that’s where the portal venous drainage went, and that the small proportion that did not metastasize to the liver were due to some anatomical differences. Our data suggests that these different patterns of metastases may instead be due to biology rather than anatomy.
You found the prevalence of HER2 amplification in pancreatic cancer to be lower than previous estimates. What are the implications of this finding?
One of the challenges in detecting HER2 amplification initially was that immunohistochemistry was difficult to interpret. Most recent findings in breast and gastric cancer suggest that detecting HER2 amplification was a better predictor of response to anti-HER2 therapies. Overestimating the degree of HER2 amplification in clinical trials greatly under-powers these trials. For example, if the true prevalence is only two percent and the measured HER2 amplification rate using other approaches, i.e. non-standardised methodologies, was over ten percent, then you would only expect at best 20 percent of patients to respond, more likely ten percent. Therefore a trial of say 100 patients would only have about ten potential responders, making it difficult to detect a positive signal in a clinical trial.
What are the challenges in utilising our growing knowledge of the different molecular characteristics of cancer in a clinical setting? How can we overcome these challenges?
We are certainly challenged in the future as we understand the vast heterogeneity of cancer in general. This is particularly important since most therapeutics are now made towards particular molecular targets. If we sub-classify cancers based on organ first and then on molecular subtype, the numbers of patients that can potentially be treated in a clinical trial becomes small and challenging to recruit. This means we simply don’t have the number of patients to test this adequately with current systems.
Furthermore, simply enlarging trial numbers becomes even more prohibitive since to detect responses in a particular sub-class that is only present in 2-3 percent requires many thousands of patients. One way that this can be overcome is using a trial model called a basket trial where patients are primarily recruited based on their molecular sub-type and then sub-stratified based on their organ of origin in novel clinical trial designs. It will always be important to classify malignancies based on organ, particularly with regard to operative resection as anatomical location is important from that perspective. Similarly histopathological classification and verification of material prior to molecular testing will always remain of paramount importance, however, we need to find ways to feasibly test this strategies in the clinic.
What’s next for your research?
The next steps in our research will be about continuing to define molecular sub-groups in pancreatic cancer, defining their particular clinical and pathological features if they are discernible, and then testing stratified/personalised medicine approaches in some of the pre-clinical models that we have developed. These pre-clinical models include patient derived xenografts and patient derived cell lines where we have fully characterised the underlying genomic, epigenomic and transcriptomic events in each particular line so that we can attempt to match the right therapy to the right molecular phenotype. Future opportunities to extend this work also exist in understanding the molecular characteristics of patients in clinical trials, analysing samples from clinical trials, and comparing responders to non-responders, with a particular focus on exceptional responders. Some of our most significant challenges lie in how we take some of these data forward, particularly in systems and paradigms that have been established for many decades. We aim to contribute to the global paradigm shift in the way we approach cancer research, translation and treatment.