Lung Nodule Risk Calculator and Cost-Effectiveness of Different Lung Cancer Screening Algorithms

See article by Hammer et al in this issue.

Brett M. Elicker, MD, is a clinical professor in the department of radiology and biomedical imaging at the University of California San Francisco (UCSF). He did a radiology residency at Yale and thoracic imaging fellowship at UCSF. His clinical and research interests are in the areas of diffuse lung disease and lung cancer.

Over the past several decades, health care costs in the United States have increased significantly and represent an ever-growing percentage of overall expenditures. It is estimated that in 1960 health care costs comprised 5% of the gross domestic product in the United States, whereas in 2016 that number increased to nearly 18%. Multiple factors contribute to this increase; however, fundamentally these can be broken down into two main contributors: greater utilization of resources, particularly in the context of an aging population, and greater cost of resources. Imaging in particular has been singled out as a major contributor to these costs (1); thus, the efficient use of this costly resource is a topic worthy of consideration and study. Efficient resource utilization is particularly pertinent for screening examinations in which imaging contributes to detecting disease in asymptomatic individuals. Screening programs often involve performing imaging examinations on large numbers of individuals; thus, improving their efficiency and cost has the potential to significantly reduce health care resource utilization and/or overall expenditures. This needs to be deftly balanced with the primary aim of screening, which is early detection of disease that has the potential to become clinically significant, thus expanding the number of quality years in patients’ lives.

Since the publication of the results of the National Lung Screening Trial (NLST) in 2011 (2), lung cancer screening has become a service that is recommended by multiple groups, but given its reliance on imaging, and more specifically CT scans, cost is a significant concern. In a study by Black et al (3), the estimated cost of screening based on the NLST results was $81 000 per quality-adjusted life year (QALY) gained. Most experts consider $100 000 per QALY gained as the threshold below which screening is considered an appropriate use of health care resources. While lung cancer screening is below this threshold, other screening programs are much more efficient. As an example, colonoscopy is estimated to cost less than $25 000 per QALY gained (4). Given that lung cancer screening is not that far below the threshold considered cost-effective, improving its efficiency is a worthy goal.

In this issue of Radiology: Cardiothoracic Imaging, Hammer et al (5) investigate the cost-effectiveness of different lung cancer screening regimens. These include the American College of Radiology (ACR) Lung Reporting and Data System (Lung-RADS) system, the British Thoracic Society (BTS) guidelines, the American College of Chest Physicians (ACCP) guidelines, and CT follow-up alone. The BTS and ACCP guidelines share in common the inclusion of nodule risk calculators in which specific clinical information is incorporated with imaging data in estimating the likelihood of a nodule representing malignancy. The specific focus of this article was the cost of Lung-RADS 4 (A, B, or X) lesions, which have the highest risk of representing lung cancer. To this end, the authors use a simulated cohort of 100 000 patients to determine the performance (sensitivity, specificity, and accuracy) and cost-effectiveness (QALY gained and cost per QALY gained).

From a performance standpoint, the BTS guidelines were the most accurate (87%) due to a relatively high sensitivity (80%) and specificity (92%). Lung-RADS was the most specific (93%) of the screening algorithms excluding CT follow-up alone, but its overall accuracy was reduced because of a relatively low sensitivity (59%). The ACCP guidelines were the second most accurate (80%) but were also the least specific (84%) because of a relatively high rate of resected nodules that were benign (17%). From a cost perspective, the BTS guidelines and Lung-RADs were the most cost-efficient in that they demonstrated the best balance between the greatest number of QALYs gained (10.041 and 10.021 years, respectively) and lowest cost ($82 362 and $81 329 per QALY gained, respectively). While the cost differences between the BTS guidelines and Lung-RADS seem small, the incremental cost-effectiveness ratio (ICER) between the two was $52 643 per QALY gained. ICER represents a comparison of the costs between two interventions in proportion to their effectiveness.

In summary, it appears that the BTS guidelines represent the most cost-efficient among all of the guidelines studied here. While the Lung-RADS system focuses solely on nodule specific features, the BTS guidelines utilize the Brock University cancer prediction equation (6), which was derived from the Pan-Canadian Early Detection of Lung Cancer study. This equation incorporates nodule specific features, but also includes clinical features such as age, sex, family history of lung cancer, and presence of emphysema. When the Brock equation predicts a greater than 10% risk of a nodule representing malignancy, PET/CT is then used as an additional triage mechanism. Subsequent workup depends on the post-PET/CT likelihood of malignancy and is divided into three categories: less than 10% likelihood of malignancy (CT follow-up), 10%–70% likelihood of malignancy (biopsy), greater than 70% likelihood of malignancy (surgical resection).

There are a variety of factors that have the potential to improve the performance and cost-efficiency of lung cancer screening. One such factor is the experience of the practitioners who treat and coordinate the care of lung cancer screening candidates. For instance, experienced thoracic surgeons are able to achieve a significantly lower mortality than general surgeons in this setting. Volume measurements of nodules reported by radiologists may be a more accurate determinant of change over time when compared with diameter measurements. Nodule risk calculators may also be helpful in maximizing the performance of lung cancer screening and minimizing the cost. In an article by White et al, the authors compared the performance of the Brock University prediction equation to the Lung-RADS system (7). In this study, they found that the Brock University prediction equation had a significantly higher accuracy in a lung cancer screening program. A variety of other nodule risk calculators exist, each having been validated on different patient populations and each with their own advantages and disadvantages.

With these facts in mind, how should the findings of the current article be applied to a real-life clinical practice? As Lung-RADS is the standard system by which most lung cancer screening studies are reported in the United States, it would seem that abandoning it in favor of another system would be premature. Both the BTS and Lung-RADS systems were cost-efficient and demonstrated relative effectiveness in maximizing QALYs gained and cost-effectiveness. Additionally, given the mandatory reporting of information on lung cancer screening to the ACR registry, there is already a vast amount of data collected using Lung-RADs algorithms. But it is important to remember that any categorization scheme, such as Lung-RADS, is unable to fully capture the complexities of each individual nodule and each individual patient. Lung nodule risk calculators should play a major role in cases such as this in which there is uncertainty about the appropriate next stage in a patient’s workup. The information yielded from a risk calculator is often helpful in conveying risk to patients and formulating, in a collaborative fashion, the next best steps in management. Patient preference in these cases often has a significant impact on management decisions, and part of that informed decision-making is quantifying the likelihood of a nodule representing malignancy.

In summary, of all the algorithms studied in this article, the BTS, which incorporates a nodule risk calculator that includes patient information, was most effective at maximizing QALYs gained and minimizing cost. Nodule risk calculators should be utilized as a complement to established lung cancer screening algorithms, such as Lung-RADS, particularly in situations in which management decisions are challenging.