CORR Insights^®: What Is the Clinical Importance of Incidental Findings on Staging CT Scans in Patients With Sarcoma?

Where Are We Now?

There is perhaps no better application of precision medicine than treating patients with cancer. For sarcomas, targeted therapies have advanced somewhat, based not on histologic phenotype, but on unique combinations of genotypic abnormalities [3]. Other emerging treatments, including checkpoint inhibition and adoptive immunotherapy, considered to be the ultimate in personalized medicine, are used with mixed results in solid tumors [8, 10]. The presence of metastatic disease in patients with soft tissue and bone sarcomas remains the single most important determining factor of long-term survival.

In the current study, Mayo and colleagues [7] investigated the natural history of “indeterminate lesions” discovered in the lung, nodes, liver, or bone when staging patients with soft tissue and bone sarcomas. The group recognized that there are only limited criteria on which to justify decisions regarding whether to observe the lesion in question with imaging studies over time or to perform a biopsy. Their goal was to identify patterns within the data that could be used to determine which of these nodules are most likely to represent metastases, and ultimately to guide treatment. In doing so, they took the first step toward applying precision medicine to this aspect of sarcoma care.

Most of the patients included had abnormalities on staging studies, and 57 of 149 (38%) of the nodules found were deemed “indeterminate.” This was a rather large proportion, in my opinion, and helps illustrate the need for better means of risk stratification. Nodules that grew, increased in number, or were confirmed histologically were considered positive for metastatic disease. At first glance, 6 months of followup seems insufficient, since patients treated for sarcomas require lifelong surveillance. However, the team did not seek to predict which patients would develop metastatic sarcoma, but rather, which of the indeterminate nodules detected during staging represented metastatic disease.

Although Mayo and colleagues acknowledged that their sample size was too small to generate a “true prediction of risk” [7], their study is both timely and interesting. If given the choice, we would all prefer to limit the amount of uncertainty, whenever possible. Identifying characteristics that could risk stratify indeterminate nodules would certainly be a step in the right direction, and perhaps help eradicate the term altogether. If developed, such a tool could result in stage migration, better known as the Will Rogers Phenomenon, where an element from one set moves to another set and, in doing so, raises the average values of both sets [2].

Unfortunately, the definition of an indeterminate nodule is rather subjective. The authors of the current study defined it as “not clearly suggestive of a specific process such as a granuloma, cyst, or hemangioma and could be representative of early metastatic sarcoma” [7]. Another group defined it as “an uncalcified, subcentimeter (less than or equal to 10 mm in greatest dimension) pulmonary nodule in the lung parenchyma” [11]. As of now, we have no universally accepted definition.

Where Do We Need To Go?

On the surface, widening the aperture to include the evidence-base from pulmonology offers no help. Well-known recommendations for the evaluation of pulmonary nodules, such as those published by the Fleischner Society [6], are not recommended for use in patients with known primary cancers. But criteria used by scoring systems designed for lung cancer screening, including age, size, location, density, consistency, and especially, specific oncologic diagnosis, may be useful as we look to apply a similar approach to a sarcoma-specific patient population [5, 12]. Importantly, unless we can agree on a strict definition for what constitutes an “indeterminate nodule,” we may be better off ignoring the term, and letting the outcomes of all patients found to have pulmonary nodules at the time of screening define risk. For this, we need more data, and a bit of collaboration.

When basing decisions on incomplete information, surgeon and author Atul Gawande MD, MPH proposed two strategies: “One is to shrink the amount of uncertainty … and agree in advance what should be done in uncertain situations” [4]. By gaining experience, each of us strives to limit uncertainty in our clinical practice. Although standardizing care across entire healthcare networks is a relatively slow process, the rate at which we “shrink uncertainty” and standardize care can, and should, be quickened as we develop precision medicine solutions. Achieving consensus for what should be done if things remain uncertain, is another matter. Where we need to go depends on if and how we, as a specialty, plan to catalyze the process of turning our clinical data into viable tools.

To be fair, a group [9] was successful in developing a scoring tool designed to risk stratify pulmonary nodules in pediatric patients, many of whom had sarcomas. However, this model does not appear to be widely used, perhaps because it remains unvalidated. This highlights an important barrier to the process of turning data into decision support tools. Any prognostic model should be updated from time to time, and most importantly, externally validated in all patient populations in which it is used. The collaborations necessary to conduct validation, and ultimately impact studies are expensive and time consuming, drawing parallels to the “Valley of Death”—a chasm between development and clinical practice, often attributed to the drug-development pathway [1].

How Do We Get There?

Where appropriate, it may be time to shift our focus from descriptive frequentist statistics to the development of precision-medicine tools. In this case, Mayo and colleagues [7] seek to estimate the likelihood of malignancy in patients with indeterminate nodules, but there are a number of other precision-medicine opportunities within orthopaedics, such as characterizing the risk of fragility fractures, infection after total joint arthroplasty, or opioid dependency after surgery, as well as characterizing treatment response or survival trajectories in patients with advanced malignancies. Prioritizing them should not be the job of the clinician-scientist, but rather of the orthopaedic societies, whose missions are to educate, arrive at a consensus, and provide guidance to the sponsors of our orthopaedic research.

By developing decision support tools, we have an opportunity to shrink some of the uncertainty we encounter in our day-to-day practice. In doing so, we may find corollaries to the implant development process. As the end-users, orthopaedic surgeons should serve as subject matter experts, while considering all available data. Rather than interfacing with biomechanical engineers, as one would if the goal were to develop an implant system, those interested in developing decision tools would interface with biostatisticians and bioinformaticians, who specialize in advanced mathematical modeling.

Just as there are no perfect implants, there are no perfect models. It is our nature to understand the indications and limitations of specific orthopaedic implants. If we are to embrace decision support, we must achieve the same level of familiarity with machine-learned models. However, we must think beyond simple indications and contraindications, which can be achieved with external validation studies. It is perhaps more important to conduct prospective outcome studies, to evaluate if and how each tool improves the care we provide.

Involvement at the society level can add some standardization and oversight. Organizing the validation and impact studies necessary to achieve academic credibility can be as complex as conducting randomized clinical trials. Societies offer experienced mentors and collaborations needed to design and facilitate them. Furthermore, there is a need to ensure consistency in the evaluation, regulation, and implementation of precision medicine tools, moving forward. Societies, the members of which already interact with regulators, policymakers, and industry, can help ensure our precision medicine solutions not only adhere to accepted standards, but also ensure the tools can be streamlined into useful orthopaedic workflows. Finally, we have an obligation to highlight the uncertainty associated with the use of precision medicine tools. Only after we understand the limitations, can we work toward incorporating them into clinical practice guidelines.