This article is more than 5 years old. Information may no longer be current.
Should a model-based approach be used to identify which individuals should undergo lung cancer screening?
Click here to read the Cover Story, “Physician, patient education needed to increase lung cancer screening rate.”
Yes.
Several major institutions that recommend lung cancer screening require or suggest that NLST enrollment criteria or similar criteria be used for determination of high risk and selection of individuals for screening. In contrast to these recommendations is the view that using accurate lung cancer risk-prediction models to select individuals for screening is superior.
The evidence is strongly in favor of using an accurate lung cancer risk-prediction model. Although the NLST criteria were efficient for enrolling individuals into a trial, they are simplistic and less efficient than using an accurate model in public health practice. A model includes more predictors and more accurately reflects their relationships to lung cancer. Multiple international studies have shown that good prediction models outperform NLST-like criteria. Good models are more sensitive at identifying individuals who develop lung cancer, and their use has been shown to be more cost-effective.
The criticism that models are hard to use is an exaggeration. The PLCOm2012 model has been focus group-tested and compared with NLST criteria. Time and difficultly of data collection were not markedly different. Administration of the PLCOm2012 for screening selection has been successfully implemented in several settings: Pan-Canadian Early Detection of Lung Cancer Study, International Lung Screen Trial, and Cancer Care Ontario’s Lung Cancer Screening Pilot for People at High Risk. In these settings, proportionately more lung cancers were identified than in the NLST.
Another criticism of models is that they select a large number of people who have more comorbidities and will die sooner from non-lung cancer causes. This is the unavoidable case, but using NLST criteria is no solution. The NLST criteria select a sizable number of individuals who have low lung cancer risk and will not benefit from screening (based on real people, real trial data). When these individuals are excluded, the remaining NLST-selected individuals have morbidity and mortality profiles similar to model-selected individuals. A safeguard exists to overcome this problem. Generally, clinicians are involved in screening programs at an entry stage and triage individuals out who have severe comorbidities and low survival probabilities.
An important reason for using selected models is reducing race disparities. African-Americans have elevated risks for lung cancer after adjusting for other risk factors. Their elevated risk is considered by some models, but not by NLST-like criteria, which leads to their underselection.
There are several good reasons to use accurate models for screening selection, and there are no strong arguments not to use them.
Martin C. Tammemägi, PhD, is professor of epidemiology in the department of health sciences at Brock University in Ontario, Canada, and senior scientist at Cancer Care Ontario. He can be reached at martin.tammemagi@brocku.ca. Disclosure: Tammemägi reports no relevant financial disclosures.
No.
Most insurance companies will not reimburse for care outside of insurance-based rules or eligibility. If an individual patient and provider go outside CMS or their insurer’s rules, both of them should make that decision with full knowledge that screening might not get reimbursed.
If it was reimbursable, there are still factors to consider. Any change in eligibility rules to use a model-based prediction is years away; however, the USPSTF is looking into this question as part of its update on lung cancer screening. Still, the task force’s recommendation will not necessarily translate to payment, because Medicare and Medicaid have their own rules. I don’t think a decision has been made, but changing the rules would be a big leap that these entities haven’t made for other tests.
Right now, a clinical provider and patient have a shared decision-making visit to discuss lung cancer screening. If that patient decides he or she wants to pursue lung cancer screening, he or she undergoes low-dose CT scan. Most likely, the individuals being offered screening by their providers are those who qualify under CMS or USPSTF eligibility criteria. Someone certainly could pay out of pocket if he or she wants. So, if a provider recommends screening for a patient based on a prediction model, cost should be included in the shared decision-making conversation.
Based on previous research and ongoing studies, the process for shared decision-making isn’t perfect, and there are gaps in what patients know about risks and benefits of screening and smoking cessation. Some published papers have suggested we are screening people who have too many comorbidities to benefit from screening. I am worried that the prediction models will overselect people with many comorbidities. American Thoracic Society published considerations about whether people who have a lot of comorbidities would benefit from screening. Most centers adhere to strict eligibility criteria, but for criteria that are more nuanced — such as those regarding comorbidities — we are probably screening more people than would benefit.
Prediction models are very important in the information exchange portion of the shared decision-making visit, but they should not be used to determine eligibility for screening. Rather, models can help share benefits and risks with individual patients. It is important to use a model to suggest to the patient in front of you what his or her benefits might be. In terms of using prediction models for lung cancer screening, we are in the “you should be encouraged to use them” phase, not the “you should definitely use them” phase.
Christopher Slatore, MD, MS, is a CIVIC core investigator at U.S. Department of Veterans Affairs and associate professor at Oregon Health & Science University. He can be reached at slatore@ohsu.edu. Disclosures: Slatore is employed by the Department of Veterans Affairs. He reports grants from American Cancer Society, NCI, Oregon Health & Science University Hospital’s Knight Cancer Institute, Patient-Centered Outcomes Research Institute and VA.