That overdiagnosis exists in any type of cancer screening is not a point of contention, as I discussed in my article . It is an inexorable consequence of our incomplete knowledge of the biological behavior of premalignant and malignant lesions and of the fundamental limitation that we cannot, by definition, predict random events. The relevant issues are how to estimate its magnitude and what solutions can be envisioned to reduce it. Measuring overdiagnosis is not a simple task, and clinical trial choices and assumptions dramatically influence its estimate .
Continuing evolution of practice guidelines as we accrue more data and better evidence will result in better risk stratification, which will likely raise size thresholds for actionable nodules, increasing specificity at the cost of slightly decreased sensitivity, thereby decreasing false positives and overdiagnosis. The first version (v1) of the American College of Radiology Lung Imaging Reporting and Data System (Lung-RADS) has already taken steps in that direction, raising the cutoff of what constitutes a “positive” screen for solid nodules to 6 mm or more versus 4 mm in the National Lung Screening Trial (NLST). Moreover, it is assumed that most overdiagnosed cancers are in the more indolent end of the spectrum of biological behavior, because these lesions have a much longer preclinical phase and contribute most to length and lead time biases. These indolent lesions have computed tomography (CT) features that are characteristic, namely, they tend to present as ground glass attenuation nodules. Therefore, radiologists can leverage this knowledge to raise the threshold of what constitutes a “positive” screen when a ground glass nodule is identified, thereby substantially decreasing overdiagnosis. LUNG RADS v1 has also accomplished that by raising the threshold for a positive screen for pure ground glass nodules to 20 mm. An analysis of the effect of LUNG RADS implementation in a clinical screening program demonstrated that these changes in threshold have decreased the overall false-positive rate from 27.6% to 10.6%, without impacting the false-negative rate. The positive predictive value was increased from 2.5% to 17.3%. More importantly, there was a virtual elimination of CT lung screening examinations positive for nonsolid nodules, which should, as discussed previously, likely dramatically decrease the overdiagnosis rate .
Every trial makes assumptions and is prone to biases. The NLST design is not flawless. However, in spite of its limitations, including relatively short follow-up duration and absence of an unscreened group, the NLST is currently the best scientific evidence that there is likely benefit of screening for lung cancer in a well-defined high-risk population. It is the largest lung screening randomized trial ever performed, and its follow-up interval of 6.5 years is one of the longest. The reduced mortality of 20% in the low-dose computed tomography (LDCT) arm was the primary reason the trial was interrupted prematurely; therefore, it is conceivable that the true benefit would be higher if the trial were continued further. The 95% confidence interval was 0.70–0.92, demonstrating statistical significance. Even if the premise that the equivalence of the computed radiography (CR) screening arm and a hypothetical null screening arm mortality does not hold, it would be statistically unlikely that the magnitude of the difference would be sufficient to invalidate the NLST results, based on analysis of the previous trial results that evaluated CR screening. A Cochrane collaboration review judged the NLST to have a low risk of selection, detection, and attrition biases in comparison to other trials . Most European trials that have released final results have failed to demonstrate similar benefits; however, these were substantially smaller, their follow-up period were shorter, and their statistical power was considerably smaller . Until a pooled analysis of all European trials is completed (European randomized lung cancer CT screening), which is expected to include seven trials and about 37,000 patients (versus the 53,454 patients of the NLST), the NLST will remain the best available scientific evidence.
A subgroup analysis of the NLST showed that in patients with chronic obstructive pulmonary disease (e.g. clinically demonstrated airflow limitation), there was a doubling of cancer incidence, no excess cancers relative to the CR arm and comparable histology, showcasing the importance of risk stratification to maximize the benefits of screening and minimizing overdiagnosis.
It is clear that major surgery such as lobectomy has costs and risks, although these vary according to the setting, with high-volume tertiary centers performing better than smaller community practices. It is imperative to avoid unnecessary surgical resections and restrict these to the patients who are more likely to benefit. Again, this can be accomplished via more conservative size and growth rate thresholds for intervention, as delineated previously. Although the current standard of care for operable, stage I, nonsmall cell lung cancer is lobectomy with hilar and mediastinal nodal sampling, there are growing data to support that minimally invasive ablative techniques, which include stereotactic ablative radiation therapy (SABR) and percutaneous ablation, will play a progressively more central role and may eventually replace surgery for early-stage nonsmall cell lung cancer. For example, a pooled analysis of data from the STARS and ROSEL trials (which were terminated early because of slow accrual) showed promising results. There was slightly better 3-year survival in the SABR versus the surgical group, and recurrence-free 3-year survival was similar. More importantly, the complication rates were substantially lower for SABR, noting that 10% of patients had grade 3 treatment-related adverse events, without any treatment-related deaths in the SABR group. In the surgical group, 44% of patients had grade 3–4 treatment-related adverse events and 4% of treatment-related deaths. These trials are small (58 patients combined) and the follow-up interval is short; however, there is suggestion that SABR may perform as well as surgery with much lower incidence of treatment-related complications. In the near future, increased utilization of this technique or less-invasive surgical techniques such as video-assisted thoracoscopic wedge resections may achieve similar local control and long-term survival with substantially decreased risks, morbidity, and preservation of lung function, when compared to lobectomy or sublobar resections . More clinical trial data are needed to prove effectiveness and efficacy of these minimally invasive modalities to eventually update practice guidelines.
Conclusion
In summary, it is crucial to understand that overdiagnosis, though unavoidable, is very difficult to measure and any estimates of its magnitude are constrained by trial design and specific assumptions. Estimates of overdiagnosis using microsimulation modeling that extrapolated NLST data to longer follow-up duration suggested an overdiagnosis rate of 6.75% in the CR arm and 8.62% in the LDCT arm , substantially lower than the 18–20% that has been suggested on other analyses. Multiple authors have proposed estimates that varied from as low as 1.2% to 5.3% , provided indolent lesions are excluded, which has been mostly accomplished by the increased size threshold of ground glass nodules for classifying an LDCT exam as “positive” using LUNG RADS criteria . It is therefore not unrealistic to ascertain the feasibility of achieving an overdiagnosis rate of 5% or less with strict adherence to high-risk criteria for screening eligibility, continuous improvement of reporting systems such as LUNG RADS with implementation of increased threshold for both solid and ground glass attenuation nodules, and better risk stratification using clinical and genomic data . Naturally, as screening is implemented in the general population in the United States, continued monitoring and analysis of the clinical data will determine whether the benefits outweigh the costs and risks, and can be used to arrive at better estimates of overdiagnosis and to further refine the strategies for diagnosis and management of screen-detected nodules, to achieve maximal societal benefit at minimum costs and harms.
Get Radiology Tree app to read full this article<
References
1. Mortani Barbosa E.J.: Lung cancer screening overdiagnosis: reports of overdiagnosis in screening for lung cancer are grossly exaggerated. Acad Radiol 2015; 22: pp. 976-982.
2. ACR : Lung cancer screening resources. Available at: http://www.acr.org/Quality-Safety/Resources/Lung-Imaging-Resources Accessed August 14, 2015
3. McKee B.J., Regis S.M., McKee A.B., et. al.: Performance of ACR lung-RADS in a clinical CT lung screening program. J Am Coll Radiol 2015; 12: pp. 273-276.
4. Manser R., Lethaby A., Lb I., et. al.: Screening for lung cancer. Cochrane Database Syst Rev 2013; CD001991
5. Heuvelmans M.A., Vliegenthart R., Oudkerk M.: Contributions of the European trials (European Randomized Screening Group) in Computed tomography lung cancer screening. J Thorac Imaging 2015; 30: pp. 101-107.
6. Young R.P., Duan F., Chiles C., et. al.: Airflow limitation and histology-shift in the National Lung Screening Trial: the NLST-ACRIN Cohort substudy (N=18, 714). Am J Respir Crit Care Med 2015; Epub Jul 22, 2015
7. Louie A.V., Palma D., Dahele M., et. al.: Management of early-stage non-small cell lung cancer using stereotactic ablative radiotherapy: controversies, insights, and changing horizons. Radiother Oncol 2015; 114: pp. 138-147.
8. Chang J.Y., Senan S., Paul M., et. al.: Stereotactic ablative radiotherapy versus lobectomy for operable stage I non-small-cell lung cancer: a pooled analysis of two randomised trials. Lancet Oncol 2015; 16: pp. 630-637.
9. Haaf K., de Koning H.J.: Overdiagnosis in lung cancer screening: why modelling is essential. J Epidemiol Community Health 2015; Epub Jun 12, 2015
10. Esserman L.J., Thompson I.M., Reid B., et. al.: Addressing overdiagnosis and overtreatment in cancer: a prescription for change. Lancet Oncol 2014; 15: pp. e234-e242.
11. Prokop M.: Lung cancer screening: the radiologist’s perspective. Semin Respir Crit Care Med 2014; 35: pp. 91-98.
12. Boeri M., Sestini S., Fortunato O., et. al.: Recent advances of microRNA-based molecular diagnostics to reduce false-positive lung cancer imaging. Expert Rev Mol Diagn 2015; 6: pp. 801-813.