For decades, we have used the same system of testing and ultimately approving new treatments for cancer. The starting point of new drugs has become sexier--by targeting genetic lesions and avoiding some of the poisonous effects of older drugs. But the rest of the sequence of trials toward the road to approval is stale. It still starts by demonstrating safety (phase 1 trials), then some sense of effectiveness in shrinking tumors (phase 2) and ultimately a large trial comparing it with "standard" treatment or to adding the new drug to standard treatment (phase 3). Drug companies are focused on meeting the metrics by the FDA for approval--usually improving survival or time to recurrence or delaying progression. This used to mean showing a statistically significant improvement in a large number of eligible patients, even if the overall improvement was small--just a couple of months of extra survival.This model is no longer working well because as we discover that cancer is broken down into small subsets that are biologically distinct, with only some subsets responding to targeted drugs. Given the side effects and spiraling costs of newer drugs, what might work for a cheaper and safer hypertension drug will certainly not apply for cancer therapies. Having to treat hundreds of patients for only a few to benefit is not sustainable, especially when newer technology is available to improve the situation. This year at ASCO, you will read our blogs about exciting updates on the cancer front, and a common theme will emerge. A much greater effort is being made to understand what subset of patients might respond even before a new agent is ready for human testing. Sometimes it is not clear what tissue tests should best qualify a clinical trial subject, so even in phase 1 testing, analysis of tissue is not built into the study in hopes of refining eligibility criteria for future trials.Another innovative trial design we use when we do not know what patient population might best respond is called a "randomized discontinuation trial." It is based on the notion that only a small fraction of patients might respond. A large number of patients are initially tested (after basic safety is established). A very small number of patients might actually have shrinkage of tumor and they stay on treatment, while those who progress are taken off the study. However, those who have neither growth nor shrinkage after a defined period of time (termed "stable disease") are randomized to either continue or discontinue therapy. If those who stop therapy exhibit tumor growth after some time more so than those who continue, it proves that the drug is effective and may even shed some light on the tumor characteristics that predict response in the small fraction of cases. Of course, the trial can be designed to allow the "randomized discontinuation" group to restart therapy with the assumption that they might derive a benefit. Another approach is called "adaptive trial design" where the results of the trial are analyzed in real time, and the randomization scheme (the fraction of patients randomized to standard vs. experiment treatment) changes over time to maximize the efficiency of the trial. While these ideas are not exactly new, they are being applied more often, especially for biologically targeted drugs since we have tools newly made available, such as rapid whole genome sequencing that can be feasibly done for at least part of the genome and are standardized to the point that they can be used for both pilot and large definitive trials. This will no doubt be a part of the major overhaul of the clinical research process currently under way in the government and drug industry sectors--the transformation of cancer clinical trials has just begun.