Henschke, Claudia I. PhD, MD
Mount Sinai School of Medicine, New York, NY
Reprints: Claudia I. Henschke, PhD, MD, Mount Sinai School of Medicine, New York, NY 10029 (e-mail: firstname.lastname@example.org).
CT screening for lung cancer has been shown to significantly reduce mortality in a high-risk population by at least 20 percent. This conclusion is based on the recently reported National Lung Screening Trial (NLST), a large randomized trial funded by the National Cancer Institute (NCI) contrasting CT screening with chest radiographic (CXR) screening. Beyond demonstrating a mortality reduction, other factors must also be considered prior to instituting a national screening policy.
As demonstrated in the United States and in Europe, guidelines are a critical component of breast cancer screening policy. Guidelines minimize potential harms by providing efficient diagnostic and therapeutic algorithms. This includes what we call the regimen of screening which specifies CT image production, radiation dose, definition of positive result, the diagnostic workup, time to diagnosis, pathologic interpretation, time to treatment and specified treatment along with integration of a smoking cessation program. The need for a multidisciplinary team using agreed upon performance indicators and audits of the performance standards are also important components.
The 1999 publication of the Early Lung Cancer Action Project's (ELCAP) study results1 showed the marked superiority of CT screening over chest radiography. The results triggered a strong public demand for information on CT screening and led ELCAP to organize the International Conference on Screening for Lung Cancer.2 An open invitation was extended to all interested in screening, including representatives from the American Cancer Society (ACS), the NCI and many other international groups and investigators. Public demand also prompted multiple NCI Advisory Board discussions and led the NCI Director to call for a Progress Review Group report that included early detection of lung cancer3; that group recommended that multiple approaches be pursued to answer all relevant questions regarding CT screening for lung cancer.
At the First International Conference on Screening for Lung Cancer in October 1999, the consensus among attendees was that pooling data from different institutions should be encouraged. ELCAP was charged with developing the protocol for such collaboration which was unanimously accepted and subsequently used for organizing NY-ELCAP4 and I-ELCAP.5 The protocol specified common elements in the regimen of screening required for data pooling, but left it to each institution to set its own enrollment criteria as to age and smoking history,6 thus broadening the knowledge base as to indications for screening. The protocol has been continuously updated at subsequent International Conferences, including development of criteria for emphysema7 and cardiovascular disease scores.8
The ELCAP design has proven efficient as relevant diagnostic information (eg, proportion of Stage I diagnoses, median tumor size) to assess annual screening can be obtained in two rounds of screening9 and the initial ELCAP results were confirmed by the NY-ELCAP.4 Once lung cancer is diagnosed, the prognostic information can be addressed using either experimental or quasi-experimental approaches. The latter was used to estimate the cure rate of 80% (95% CI: 74%-85%) for lung cancers diagnosed by CT screening among the study's 31,467 participants following the I-ELCAP protocol.7 In the absence of screening, the cure rate was about 10%. Thus, a substantial gain in the cure rate was demonstrated and the possible biases of the estimate were also addressed.
Motivated by the ELCAP results, the NCI started the NLST in 2002 to compare CT with CXR screening. The NCI study used 25,000 high-risk participants for each arm of the trial and provided 3 rounds of screening per arm. CXR screening was the control arm because another randomized trial started in 1993, the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO), compares CXR with no screening. On November 4, 2010, the NCI reported that the null hypothesis had been rejected as a 20.3% mortality reduction had been reached.
How do we reconcile the difference in the I-ELCAP estimated cure rate of 80% and the NLST mortality reduction estimate of 20.3%? The answer lies in understanding they are measuring very different parameters and even though they appear quite different, they are consistent with each other. Clearly, the two measures are inter-related; there cannot be a reduction in mortality unless some participants with potentially fatal lung cancer had been cured by the early diagnosis provided by the CT screening followed by early treatment. But why the big difference?
I-ELCAP estimated the cure rate which is a quantitative measure that is dependent on the regimen of screening which includes the low-dose CT scan. The NLST tested the null hypothesis of no morality reduction of the low-dose CT scan when compared with CXR against the alternative hypothesis of a mortality reduction of at least 20% due to the CT scan. The design uses a limited number of rounds of screening and years of follow-up to reduce the time and cost of the trial. Importantly, the magnitude of the mortality reduction for the alternative hypothesis is an arbitrary value that the designers set. Whether a statistically significant difference between the two hypothesis can be found depends on the design parameters, the actual diagnostic workup and treatment that was performed in the trial and of course, the initial low-dose CT scan provided in each round of screening. Once significance is reached, the trial is stopped.
The full mortality reduction provided by CT screening would only be reached if: screening is continuous for many years, probably 10 years; a well-defined, optimal protocol (eg, I-ELCAP3) is used; and, the analysis focuses on the appropriate time interval where the maximum mortality rate reduction becomes manifested, some 7 to 10 years after baseline, and a control arm which provides no screening is used. These requirements are illustrated by analysis of a CT screened cohort compared to a cohort with no screening10 and, likewise for breast cancer9 and other cancers as well.
The profound difference in the study design needed to provide a quantitative estimate and one used to test of hypothesis is often not fully appreciated, but should be!
1. Henschke CI, McCauley DI, Yankelevitz DF, et al. Early lung cancer action project: overall design and findings from baseline screening Lancet.. 1999;354:99–105
2. International Early Lung Cancer Action Program Conferences, Consensus Statements, and protocol. www.IELCAP.org
4. New York Early Lung Cancer Action Project Investigators. . CT Screening for lung cancer: diagnoses resulting from the New York Early Lung Cancer Action Project Radiology.. 2007;243:239–249
5. International Early Lung Cancer Investigators. . Survival of Patients with Stage I lung cancer detected on CT screening NEJM.. 2006;355:1763–1771
6. Henschke CI, Yankelevitz DF, Smith JP, et al. Screening for lung cancer: the early lung cancer action approach Lung Cancer.. 2002;35:143–148
7. Zulueta J, Wisnivesky JP, Henschke CI, et al. Scoring of Emphysema detected on low-dose CT Predicts Death from Chronic Obstructive Pulmonary Disease and Lung Cancer. Submitted.
8. Shemesh J, Henschke CI, Shaham D, et al. Ordinal scoring of coronary artery calcifications on low-dose CT scans of the chest predicts deaths from cardiovascular disease Radiology.. 2010 In press.
9. ELCAP Investigators. . An update of CT screening for lung cancer Seminars in Ultrasound, CT, and MRI.. 2005;26:438–356
10. Henschke CI, Boffetta P, Gorlova OY, et al. Assessment of lung-cancer mortality reduction from CT screening Lung Cancer.. 2010 Dec 16. [Epub ahead of print].
© 2011 Lippincott Williams & Wilkins, Inc.