The NLST Puzzle

Three years ago, the National Lung Screening Trial[1] (NLST) reported a 20% reduction in lung-cancer mortality with annual low-dose CT (LDCT) screening. The NLST findings were seminal: the first demonstration that early detection could reduce mortality from the cancer that’s the world’s leading killer.

Uptake of LDCT screening has been slow, but with the recent grade B recommendation[2] of the United States Preventive Services Task Force (USPSTF) for LDCT screening, which mandates that private insurers cover it without cost sharing, the time might be right for LDCT screening to expand considerably. Therefore, it is now critical to examine how to translate the results of the NLST (and other LDCT studies) into the clinical-practice setting.

As a so-called efficacy trial, the NLST was designed to determine whether, under optimal conditions, LDCT screening could reduce lung-cancer mortality. It was not designed to answer the question of how LDCT screening should be implemented in the clinical-care setting, nor how well it would perform there. A cancer-screening trial is a large, long-term, and expensive instrument designed to answer a single critical question.

In contrast, myriad questions need to be addressed concerning the population implementation of LDCT screening. Who should be eligible for screening, in terms of age and smoking history? How frequently should screening be performed? How should a positive screening result be defined? Should a standard diagnostic follow-up algorithm be employed—and if so, which one? What training is required for radiologists, and what types of institutions should be allowed to perform LDCT screening?

The first of these questions has received some scrutiny recently. By considering it in more detail, one can get some sense of what is involved in translating trial results into clinical practice. The USPSTF recommendations deviate slightly from the NLST eligibility criteria (age 55 to 74 years; a smoking history of 30 or more pack years; and having stopped smoking, if they no longer smoke, within the past 15 years) by modestly expanding the upper age limit to 80 years.[2]

The National Comprehensive Cancer Network[3] expanded the NLST criteria by including people with 20 or more pack years of smoking history with other lung-cancer risk factors, including occupational exposures and history of chronic obstructive pulmonary disease. Other organizations’ guidelines, such as those of the American Cancer Society[4], generally follow the NLST criteria.

It is useful to conceptualize the question of extending LDCT screening beyond the NLST–eligible population by dividing it into two components. First, is the efficacy of screening, as measured by the percentage of reduction in lung-cancer mortality, likely to differ in an expanded (lower-risk) eligible population? Second, assuming equivalent efficacy, is the benefit-to-harm trade-off worthwhile in this expanded group? The harm of LDCT screening includes false positives (with the resulting diagnostic work-up and complications thereof), as well as radiation exposure.

Efficacy Drivers

The primary drivers of LDCT efficacy are the characteristics of incident lung cancers: A similar cancer profile in the expanded (versus NLST-eligible) population should result in similar LDCT efficacy. Lowering the pack-year requirement to 20 years, for example, or increasing the time since smoking cessation to 20 years would be unlikely to change the histology (or other characteristics) of the lung cancers appreciably, and many cancers would still be caused by cigarette smoking.

Therefore, similar LDCT efficacy would be a reasonable assumption for a modestly expanded eligibility group. An NLST subset analysis[5] found no evidence of differential LDCT efficacy by smoking status (current versus former), even though the groups differed by a factor of two in lung-cancer–incidence rates.

For the second component, the benefit-to-harm ratio for screening is typically measured as the number of people who must be screened to prevent one death from the relevant cancer.[6] This figure, number needed to screen (NNS), is computed as the reciprocal of the difference in lung-cancer mortality rates among the screened and unscreened populations; for the same efficacy (percentage reduction), halving the lung-cancer risk in the unscreened population doubles the NNS.

Overall, the NLST showed an NNS of 320 for LDCT.[5] There were, however, substantial differences in the NNS between trial subgroups. For current versus former smokers, there was no evidence of differential efficacy, and the estimated NNS was twice as high for former (460) as for current (230) smokers.[5] For the population (age 55 to 74) of current smokers (or those who stopped within the past 15 years) with under 30 pack years, the NNS was estimated at about 700.[7]

It is difficult to determine the level of harm-to-benefit trade-offs at which LDCT screening is still worthwhile—from the societal standpoint, in terms of cost effectiveness and resource allocation, as well as from the patient and provider standpoints. Some[8] consider even the NLST criteria too broad, with a fraction of the NLST-eligible subjects having too low a cancer risk to justify screening. Others[9] would expand eligibility to those at moderately lower (but still elevated) risk, believing that the benefits of preventing lung-cancer deaths would still outweigh any harm. It might take years of LDCT population screening before a consensus emerges on this issue.

In practice, it might be difficult to verify subjects’ smoking history (pack years and, to a lesser extent, time since smoking cessation), since these might be documented in medical records only rarely. Therefore, some intermediate-risk subjects interested in being screened (and having screening covered by insurance) might well embellish their smoking histories, thus rendering moot the question of how to define an optimal smoking-history criterion.

Although another randomized trial of LDCT is unlikely to be conducted in the United States, ongoing research is clearly needed to determine how to implement LDCT screening optimally in this country. Therefore, it will be critical to monitor how such screening is carried out in diverse practice settings and to analyze how implementation parameters correlate with the various outcomes, both immediate and downstream, of LDCT screening. Such exercises will not provide the definitive answers of a randomized trial, but they will serve as guides to best practices for population LDCT screening.

Paul Pinsky, PhD, MPH, is acting chief of the Early Detection Research Group in the Division of Cancer Prevention of the National Cancer Institute.

References

1. Aberle DR, Adams AM, Berg CD, et al. Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med. 2011;365(5):395-409.
2. United States Preventive Services Task Force. Screening for lung cancer. Published December 2013. Accessed April 11, 2014.
3. National Comprehensive Cancer Network. NCCN guidelines. Accessed April 11, 2014.
4. American Cancer Society. American Cancer Society guidelines for the early detection of cancer. Published May 3, 2013. Accessed April 11, 2014.
5. Pinsky PF, Church TR, Izmirlian G, Kramer BS. The National Lung Screening Trial: Results stratified by demographics, smoking history, and lung cancer histology. Cancer. 2013;119(22): 3976-3983.
6. Rembold CM. Number needed to screen: development of a statistic for disease screening. BMJ. 1998;317:307-312.
7. Pinsky PF, Berg CD. Applying the National Lung Screening Trial eligibility criteria to the US population: what percent of the population and of incident lung cancers would be covered? J Med Screen. 2012;19(3):154-156.
8. Kovalchik SA, Tammemagi M, Berg CD, et al. Targeting of low-dose CT screening according to the risk of lung-cancer death. N Engl J Med. 2013;369(3):245-254.
9. Goldberg P. Conversation with the Cancer Letter: Kazerooni: the case for broad coverage. The Cancer Letter. Published March 21, 2014. Accessed April 11, 2014.