有些道理
...............,
H. Jack West MD, Medical Director, Thoracic Oncology Program, Swedish Cancer
Institute, Seattle, Washington
In an effort to expedite the transition into a world of molecular oncology
and precision medicine, the US Food and Drug Administration (FDA) recently
approved Foundation Medicine's next-generation sequencing (NGS)-based in
vitro diagnostic test, FoundationOnc CDx (F1CDx), which can detect genetic
mutations in 324 genes and two genomic signatures in any solid tumor type.[1
] Moreover, the Centers for Medicare & Medicaid Services (CMS) also proposed
coverage for the F1CDx test.[1] We can certainly expect this to lead to a
transformation in cancer care and a greater frequency of detection of
potentially actionable targets. At the same time, however, it is important
to recognize the limitations and potential negative, unintended consequences
of this change as well.
In the interests of full disclosure, I have previously expressed skepticism
about universal genomic sequencing,[2] on the basis of the absence of
prospective, population-based evidence that genomic testing improves any
clinically significant outcomes. This is not to say that molecular testing
has no role, particularly for patients with cancers for which we have well-
established and biomarker-dependent targeted therapies. My concerns about
universal broad genomic testing have not intended to question the utility of
"precision medicine."
Instead, my view is that we should step back from the unfettered hype to
question the presumption, still based on anecdotal or very biased
retrospective data combined with an inherent belief or bias in favor of NGS
testing, that identifying more targets is inherently better, especially if
most of these identified targets have little or no clinical data to guide an
arguably optimal targeted therapy toward them.
But regardless of whether or not there is a benefit to identifying rare
mutations with potential but unproven therapies, NGS testing represents a
more efficient use of limited tissue and becomes increasingly cost-effective
as the number of targets clearly worth testing for a given cancer grows.
For lung cancer, where EGFR, ALK, ROS1, and BRAF B600E are all indicated
tests in addition to PD-L1 expression, and with another handful of emerging
biomarkers for associated targeted therapies for which there is a critical
mass of phase 2 clinical evidence, NGS testing is a clearly compelling
option. For other cancers, in which there are few or no current molecular
markers with established utility, the benefit of testing for a wide range of
markers is more speculative.
There is good reason to be concerned that more is not necessarily better in
this setting. Although the cost of broad NGS testing itself is often cited
as a significant concern, this expense of several thousand dollars per test
is little more than a drop in the bucket in a time when we have become
inured to treatment costs often exceeding $15,000 per month. This is
especially true as technology improves and NGS testing costs may well
decline. Our bigger concern should be the financial and clinical
consequences of treatment decisions made for molecular targets without an
associated targeted therapy.
Although many respected leading figures in oncology and healthcare presume
that treatments based on molecular testing will lead to better outcomes,
this conclusion is devoid of prospective evidence. Dr William Halsted was a
renowned, commanding figure who fostered the eminence-based dogma that
disfiguring radical mastectomies were the appropriate treatment for
resectable breast cancer, and this edict was broadly accepted as self-
evident rather than subjected to critical testing for decades—to the great
harm of generations of women with breast cancer. Unfortunately, even after
centuries of what should be humility-inspiring illustrations of our hubris,
the medical community remains all too willing to grant a pass to bias-driven
conclusions from respected luminaries who speak with authority but weak
evidence. Precision medicine needs to move beyond dramatic anecdotal cases
without denominators of the population tested, or retrospective studies
working backward from collections of patients with readily targetable
lesions.
Despite the reflexive promotion of broad NGS testing as the party line by so
many of its proponents, the fact remains that a large fraction of potential
biomarkers identified by NGS testing create dark tunnels that are just as
likely to be rabbit holes leading nowhere as they are to be previously
unrecognized gold mines of remarkably beneficial and effective therapies for
new or future targets. In an effort to provide treatment options and
justify the cost of testing that all too often delivers no evidence-based
therapy recommendations, NGS reports from Foundation Medicine and many
others offer treatment options based on very early clinical trials or even
preclinical data. Whether broad genomic testing becomes a beneficial
inflection point for improving cancer care or a key moment where treatment
costs come off the rails depends entirely on whether the reported results
are interpreted judiciously or with an uncritical, gravitational pull toward
molecularly defined therapies.
At a time when we face difficult decisions about how to cover the societal
costs of very expensive cancer treatments with well-established benefits, we
should be wary of increasing these challenges by selecting very expensive
therapies devoid of strong clinical evidence. A few months of completely
speculative, low-yield treatment with a targeted therapy that costs tens of
thousands of dollars on the basis of a preclinically driven premise has the
potential to create a societal financial nightmare if we expect these costs
to be borne by insurers, who will invariably pass the costs on to their
subscribers, now unwittingly forced to bankroll ill-advised decisions.
But the potential harm is not merely financial and indirectly borne by
society. First, these targeted therapies may have significant adverse
effects and lead to unexpected detrimental effects. Nobody in the lung
cancer community expected gefitinib to be associated with a significantly
detrimental effect after chemotherapy and radiation that amounted to a 12-
month shorter survival than with placebo,[3] even if we might have thought
that it may or may not help. But even beyond the potential for harm from a
targeted therapy compared with placebo or supportive care, I too often see
colleagues displace standard-of-care, established treatments with ones
inspired by NGS reports indicating benefits that should be considered more
in the range of "fanciful" than "expected"-a poor trade that may lead to
worse outcomes rather than better ones, especially if this decision is made
in an earlier line when treatments with meaningful and proven benefits are
bypassed.
Despite these concerns, I recognize the great potential for broad genomic
testing to accelerate the rate of cancer research and identify subsets of
patients likely to benefit profoundly. The key lies in the ability of
clinicians to discriminate the quality of the evidence and judge the merit
of the therapeutic recommendation accordingly. An important model of this
approach has been published[4] and distinguishes among treatments that are
ranked from level 1 (FDA-approved) to level 2 (standard care), level 3 (
clinical evidence), and level 4 (biological evidence). Following the
threshold for these criteria makes good sense in helping to distinguish what
to expect to be covered by insurers compared with other options. Those
options at a higher level, with relatively weaker evidence, may still
present a strong strategy in the setting of a clinical trial, compassionate
use program, or self-payment for a targeted therapy; but they should be
acknowledged as falling short of the standards that should compel insurance
coverage or substitution for treatment options with clinical data
demonstrating a benefit in treatment outcomes.
Of note, we must also maintain rigor in our endpoints and demand that the
benefits measured are clinically meaningful, such as a significant
improvement in overall survival, rather than manufacturing new, weak
endpoints-for example, a relative gain in progression-free survival compared
with ineffective alternative salvage therapies, which have been created
merely as a very low bar for tested therapies to exceed. For the cost of
these treatments and the promise we have offered from precision medicine
approaches, we should test endpoints that do more than create meaningless
feel-good victories.
Rather than continuing to debate whether broad genomic testing should be
done for patients with solid tumors, I feel that ship has sailed. But
whether it takes us to a new world or we end up under water depends on how
we direct that ship from here.
References
