Over the past 20 years, we have witnessed an unprecedented development of new life-science technologies. Although clinical outcomes have also improved, the benefits often remain out of reach for many patients. Translating scientific and technological gains into clinically meaningful outcomes that are accessible and affordable to all who need them is one of the great challenges of our time.
Several prominent voices have called on the field to “raise the bar” and aim higher in research efforts, to measurably and meaningfully lengthen and improve patients’ lives. The American Society of Clinical Oncology (ASCO), for example, has defined minimum thresholds for “clinically meaningful outcomes,” such as three-months additional survival for patients diagnosed with advanced non-small-cell lung cancer. The crucial question, though, is how to achieve these gains.
We suggest a pragmatic approach: Acknowledge the challenges facing drug-development research today and pursue effective long-term strategies. We highlight three realities that are often overlooked and discuss ways to address them.
Reality #1: Headline results from initial clinical trials do not predict the ultimate value of a treatment
A systematic analysis based on the Continuous Innovation Indicators (CII) tool (submitted for publication) shows that effect sizes obtained in registration trials have never predicted the ultimate success of the treatment. Most treatment staples used today entered the market with modest effects and the use of surrogate endpoints. Their full value developed over decades of clinical use.
Figure 1 shows the hazard ratio for overall survival—the ratio of survival in the treatment group to survival in the placebo group—of all positive studies of systemic therapies in the CII over time (n = 224). There is no relationship between hazard ratio and year of publication (p = 0.6). The average hazard ratio has remained remarkably steady at around 0.7 since it became commonly reported in the mid-1990s.
This reality raises a formidable challenge: if the hazard ratio of overall survival—the primary outcome measure from most late-phase clinical trials—is not by itself a reliable indicator of a treatment’s long-term benefits, what additional metrics can we use to assess new treatments?
Figure 1. Hazard ratios for clinical trials of cancer treatments over time
Reality #2: Drug development has become more expensive
The overall costs of drug development continue to increase. The publishing of the human genome in 2000 dramatically accelerated drug development by revealing the blueprint of every molecular drug target. Soon after, monoclonal antibodies (the blue line in Figure 2), a new class of highly specific anti-cancer molecules, proved efficacious in a wide variety of different diagnoses and applications.
Importantly, this new molecular understanding of disease transformed common cancers, like lung cancer, into collections of distinct but related diseases. Molecularly targeted drugs have specific therapeutic mechanisms, often aimed at just one molecular subtype. Thus, clinical trials that used to enroll “patients with lung cancer” must now seek patients with much rarer subtypes. Identifying enough patients requires operating more clinical sites with greater associated costs.
Figure 2: Evidence-Scores (E-Scores) as a measure of progress in 4 domains of cancer treatments classified in the Anatomical Therapeutic Chemical Classification System (ATC)
Note: Milestone symbols indicate critical historical events.
The pivot toward personalized medicine led to dramatic changes in drug development. Around 2006-2007, many observers wondered if major pharmaceutical companies would survive the shift from blockbuster drugs toward individualized treatments. Meanwhile, over the past decade, many national governments and third-party payers have tacitly adopted a threshold of approximately $50,000 to $100,000 as a proxy for determining the value of an additional year of life. Regardless of whether these thresholds are reasonable for common diseases, payers have traditionally been willing to pay substantially higher prices for rare diseases, in part because the rare nature of these conditions mitigates the overall budget impact.
The past ten years have thus set the stage for the “perfect storm” in health economics: increased molecular understanding of common diseases has turned the “big killers” into molecularly defined subtypes of cancer, each of which is rare in the overall population and responds to an individualized treatment algorithm. The industry now faces a key new challenge to make personalized drugs for fewer and fewer patients available at blockbuster prices — that is, at prices that provided an adequate return on drugs with much larger patient populations.
Reality #3: A small trial is more likely to be a false positive than a large trial with the same nominal level of statistical significance
We often say that the results of a study are “statistically significant” if the p-value is less than 0.05. But what does this really mean? Applying a p-value threshold of 0.05 means that we are willing to accept false positive results no more than 5 percent of the time. By conducting more and smaller trials, we will increase the total number of false positive findings. For example, instead of running one large trial with 4,000 participants, a sponsor can run 20 small trials with 200 participants each. One of these smaller trials is likely to return positive results just by chance even if the treatment is completely ineffective.
Pharmaceutical companies have conducted very large trials for many decades for good reasons: Larger trials indeed offer greater certainty, while small trials produce more false-positive results (“noise”). Because these false-positive trials may have large nominal effect sizes, they risk consuming large amounts of resources and set the bar unreasonably high for later treatments. Large, well-powered trials provide greater certainty that the observed effect is not just a fluke and will be replicable in other studies and in the real world. Whether a measured outcome will be clinically relevant in practice has, historically, taken many decades of real-world use to determine.
Nonetheless, we must acknowledge that increasing molecular classification of disease will, in many cases, necessitate smaller trails due to the relatively small number of eligible patients. This poses a great challenge to design clinical trials with statistically robust results for small patient populations. Smaller trials are thus not a panacea to finding clinically relevant outcomes. They are a necessity in an environment that pivots toward rarer diseases, and they will require sophisticated and rigorous statistical evaluations to deal with the increased noise in the data.
Toward a long-term solution
In this difficult environment, it is critical for all stakeholders to collaboratively identify and support long-term solutions. A plethora of new treatments is entering the market, and it will take many years for us to measure their true value. Yet the costs of these treatments substantially burdens health care systems today.
To cope with these challenges, we must raise bars. But which are the right bars to raise?
Bar #1: Empowered patients and informed physicians
Patients need to be more involved in informed decision making, and doctors need more resources to support them. Cognitive computing algorithms that are increasingly used to support clinical decision makers by breaking down complex patterns into actionable pathways represent an important step forward. ASCO supports one ambitious effort through its CancerLinQ tool.
Bar #2: Better business models
The era of personalized medicine will demand different business models that reward innovation while making lifesaving therapies available to the patients who need them. Although stakeholders have recognized the need for new business models for at least a decade, little has changed. Value-based contracts, risk-sharing agreements, and indication-specific pricing are promising ideas, but political and economic barriers have prevented widespread adoption. Pharmaceutical companies, public and private payers, and policymakers must work together to overcome the inertia of the status quo and chart a sustainable course forward.
Bar #3: Innovation in science and statistics
We need to provide sustained support for basic science to identify better targets in the pre-competitive domain. This will accelerate the development of new and more effective therapies.
Scientists and statisticians must also work together to develop new clinical trial designs that are well-suited for producing robust results with fewer patients.
As long as payers, providers, regulators, and manufacturers disagree on the realities described above, stakeholders will continue to struggle to work together productively. Identifying the critical challenges and utilizing all available evidence will be essential for effective advocacy for more resources for basic research, sustainable business models for personalized medicine, and additional resources to help patients and their doctors make informed choices about their care.
The challenge we face is complex, but it is not insurmountable. As Figure 2 shows, the field has done well so far translating genomic insights into new treatments. We urgently need durable solutions to make these gains affordable to all.