Multiple myeloma is a complex cancer, different in every patient. This complexity makes it difficult to predict what therapy will work best for each patient. Knowing a patient’s prognosis at the time of their diagnosis can help their care team decide which therapy would be most effective for first line treatment.
Differences in myeloma in each patient can be traced back to the complex DNA changes (also known as the “genomic landscape”) in each patient’s myeloma cells. This information can only be measured in tumor cells collected during the bone marrow biopsy. Researchers have long looked to catalog what these changes are, and which ones are important to a patient’s overall survival. This type of work requires a large dataset of DNA sequences from many myeloma patients, their clinical history (i.e., what they were treated with and how well the treatment worked) and their demographic data. No single clinical center in the world sees enough patients to generate a data set like this on its own.
We began our multi-year CoMMpass study back in 2011 to build such a dataset. It is the largest and most complete dataset of its type in myeloma, and more than 605 researchers around the world have used this dataset to form and test their hypotheses. CoMMpass holds the detailed genomic and clinical data of over 1000 newly diagnosed patients from 76 centers in 4 countries. These patients agreed to join this research study, enrolled over the course of 4 years and were followed for 8 years; the data is freely available to researchers worldwide. Numerous papers are published yearly using CoMMpass data, and it has been invaluable in helping to develop early risk and prognosis models in myeloma patients.
Recently, Dr. Francesco Maura of the Sylvester Comprehensive Cancer Center at the University of Miami, and colleagues from various institutions, published a new model of genomic classification and prognosis in multiple myeloma. The model was developed using detailed clinical, genomic, and therapeutic data from 1933 patients, 1062 of which were from the CoMMpass dataset. Using this data, the investigators developed a model called Individualized Risk Model for Myeloma (IRMMa) that found 12 genomic groups with high prognostic accuracy compared to earlier models. The IRMMa model was then confirmed on 256 patients enrolled in the GMMG-HD6 clinical trial and was used to show which patients benefitted significantly from autologous stem cell transplant (ASCT) vs. those who saw limited benefit. The model performed well, with good overlap seen between observed risk and risk predicted by the model for each patient. This machine learning model is now publicly available for other myeloma researchers, who are encouraged to add their own patient data to it. Further testing is necessary before this preliminary model becomes a clinical test. It is only through analyzing large data sets that researchers can accurately identify predictive signatures that could guide clinical decision-making.
In conclusion, predictive models of this type require large datasets for their accurate development. The MMRF built the CoMMpass dataset for exactly this purpose, and this treasure trove of myeloma data is now bearing fruit. It is hoped that additional data added to the IRMMa model will sharpen its predictive ability so that eventually, every newly diagnosed patient will receive the best, most effective therapy for their own unique type of myeloma.