Precision and uncertainty
Click Here to Manage Email Alerts
I am currently reading a fascinating book — The Perfectionists: How Precision Engineers Created the Modern World, by Simon Winchester.
The book describes how increasing precision has been an important driver of technological advancement. Winchester explains this largely based on the progress of the Industrial Revolution, citing examples such as the development of the steam engine through to the jet engine; the early origins of optics and camera design culminating in the precision required for the Hubble Space Telescope; and the contrasting requirements for precision in the automobile industry, including the high-level engineering in a Rolls-Royce compared with assembly-line vehicles such as the original Ford Model T.
Winchester also highlights some of the potential pitfalls of increasing precision and perfectionism. In an industrial context, this includes the social disruption and loss of jobs that result from the increased automation necessary to ensure precision and consistency. It also includes the decline of craftsmanship and creativity.
If the reader should question what could possibly go wrong with the increasing quest for precision, the author cites the example of a near-catastrophic jet engine failure on a commercial aircraft resulting from an engine component that was 0.5 mm too thin — the increasing requirement for precision also reduces the tolerance for error.
Precision in oncology
The world of oncology is similarly driven, maybe even dominated, by a trend toward increasing precision. The last couple of weeks have brought this home to me in several domains of cancer care.
Earlier today, I took part in a meeting to discuss our current next-generation sequencing platforms, and how we handle those data from the perspectives of interpretation, decision support, clinical trial eligibility and research.
When most of us think of precision oncology, we probably think first of tumor sequencing to identify actionable mutations to direct the use of targeted agents, or germline sequencing to identify individuals at elevated risk for certain cancers. A core dilemma regarding the choice of panels is essentially one of precision — do we apply narrow panels directed at a few genes of interest, or broader panels that explore many more genes, with increased precision, accepting that this approach will generate data that are currently not actionable?
The tradeoff for this increased redundancy of the more precise panel is availability of research data, plus the fact that as new targeted agents emerge, patients who might benefit can be identified from existing databases. The challenge is to determine whether the increased resources needed to collect, interpret and curate these more precise panels has true value compared with a more pragmatic panel chosen for its current applicability.
An article published last month in JAMA describes a similar situation with respect to direct-to-consumer genetic tests. Some of the currently available tests lack the precision to detect important variants in genes such as BRCA1 and BRCA2 — they lack the precision required to be confident that a negative test result is truly negative. On the other hand, as precision increases with larger panels, variants of uncertain significance emerge and the clinical decisions regarding these get blurry.
The trend toward increased precision has also generated uncertainty in treatment of localized cancer. The use of robotic surgery continues to expand in many contexts. There are well-documented improvements in some outcomes, such as postsurgical recovery times, for several diseases, but there is ongoing controversy about the true benefit in other diseases, particularly with respect to disease control and survival outcomes.
The conflict is similar for radiation therapy of localized or oligometastatic disease. There are now well-documented examples of the benefit of stereotactic radiation for patients with good control of their primary tumor. Whether the increased precision achieved with proton beam therapy — and more recently with other heavy ions — improves outcomes is less certain for most indications. The use of more precise beams can clearly be shown to limit normal tissue damage but, except for some pediatric tumors and rare cancers in adults, including brain tumors, there is limited evidence that this increased precision has meaningful clinical benefit.
Increasingly, precise local therapies also demand increasingly precise imaging and other detection modalities to ensure that accurate tumor margins are identified. As with the jet engine example, the tolerance for error decreases as precision increases, and the risks to surrounding normal tissues must be balanced against the risks for undetected tumor outside the treatment margins.
The scope of “precision prediction” also has increased in recent years — the ability to use patient and tumor characteristics to predict prognosis with some accuracy has increased and factors into many treatment decisions for patients with diseases such as breast cancer. As real-world data sets expand and artificial intelligence capabilities grow, clinical decisions increasingly will be made based on highly precise predictive algorithms.
Blending art with tools
Changing gears completely, one area in which precision seems to have eluded us — or at least me — is in prediction of the need for, and design of, future cancer services.
At my institution, we will be expanding our clinical space and are working to predict future demand for our services, future volumes and the exact configuration of the space.
I have experienced similar processes at other centers and am struck by how uncertain this is. Although we can make reasonably accurate predictions about incidence and volumes based on demographics and incidence trends, we can’t predict for changes in oncology practice that might profoundly affect our services. As an example, the often predicted and anticipated decline in the number of inpatient beds needed for patients with cancer hasn’t materialized — if anything, the requirement has increased. Sometimes, uncertainty wins out over precision.
I see definite parallels between the ascent of precision in industry and in oncology, with both positive and negative implications.
High precision probably does have diminishing returns — evidence from very detailed genomic panels and from increasingly precise surgery and radiation suggests there is a point at which further refinement has little clinical benefit. In the same way that increased automation in industry resulted in a smaller and often less skilled workforce, reliance on precise predictive tools may de-skill some of those taking care of patients — how many of us have heard our colleagues lament, “We won’t need oncologists anymore, just computers?”
Some examples from industry teach us that the blending of precision engineering with true craftsmanship produces incredible results.
As long as we continue to blend the art of oncology with the increasingly precise tools we are getting, the future for our patients will be brighter — assuming, of course, that we can do this in a way that is affordable.
References:
Kilbride MK and Bradbury AR. JAMA. 2020;doi:10.1001/jama.2019.22504.
Winchester S. The Perfectionists: How Precision Engineers Created the Modern World. New York, NY; HarperCollins Publishers; 2018.
For more information:
John Sweetenham, MD, FRCP, FACP, is HemOnc Today’s Chief Medical Editor for Hematology. He also is associate director for clinical affairs at Harold C. Simmons Comprehensive Cancer Center at UT Southwestern Medical Center. He can be reached at john.sweetenham@utsouthwestern.edu.
Disclosure: Sweetenham reports no relevant financial disclosures.
Collapse
Click Here to Manage Email Alerts