NLST

Published

Berg CD, et al. "Reduced lung cancer mortality with low-dose computed tomographic screening". The New England Journal of Medicine. 2011. 365(5):395-409.
PubMed • Full text • PDF

Clinical Question

Among patients at high risk for lung cancer, do low-dose screening CT scans reduce mortality when compared to screening CXRs?

Bottom Line

Among patients at high risk for lung cancer, low-dose screening CT scans reduce lung cancer mortality when compared to screening CXR.

Major Points

Until recently, no major practice guidelines have supported lung cancer screening among the general population. However, as lung cancer leads among the cancer-related deaths in the US, several studies have investigated the role of screening high-risk populations. Several early studies compared CXR to usual care without success. The controversial I-ELCAP trial (2006) successfully detected early lung cancers in smokers using CT scanning, although it lacked a control group. Following the development of low-dose CT scanning using one-fifth of the usual CT radiation, trials like LSS (2005) and DEPISCAN (2007) demonstrated feasibility of a large screening trial with low-dose CT using CXR as control.

Published in 2011, the National Lung Cancer Screening Trial (NLST) randomized 53,454 adults at high risk for lung cancer to annual screening with either low-dose CT or CXR. It was terminated early after a 2010 interim analysis demonstrated that at a mean follow-up of 6.5 years, low-dose CT resulted in a clear survival benefit over CXR. The authors calculate that the NNT to prevent one lung cancer death at 3 years is 320. Moreover, they suggest that 7 million Americans meet criteria for the NLST protocol, which is estimated to add $1.3-2 billion to the national health care expenditures annually; this amounts to preventing some 8,100 lung cancer deaths at a cost of $240,000 to prevent each death. In comparison with other well-accepted screening modalities such as pap smear, colonoscopy, and mammogram, low-dose screening may cost less per life-year saved.

Although the trial was generally well received, it was met with several criticisms. Chief among them was the rate of false positives with low-dose CT, resulting in unnecessary follow-up invasive testing. If NLST-type screening was employed using modern scanners with higher resolution and thus higher sensitivity for cancer detection, it would likely result in even greater rates of false positives. Related to this, there was a trend towards overdiagnosis -- in which clinically occult or insignificant cancers are detected -- among patients in the low-dose CT screening arm (1060 vs. 941).

The authors subsequently developed and validated a risk-prediction model for lung cancer utilizing age, sex, race, BMI, PYH, years since cessation, emphysema, and family history of lung cancer. Those in the NLST cohort with the highest risk for lung cancer had the greatest mortality reduction and lowest rate of false-positive screenings.

Guidelines

USPSTF lung cancer screening (2013, adapted)

• Recommend annual screening with low-dose CT if age 55-80 and 30 PYH and current smokers or those who have quit in the past 15 years (grade B)
• Discontinuation of screening once a person hasn't smoked for 15 years, limited life expectancy, or ability/willingness to have curative lung surgery (grade B)

American Cancer Society lung cancer screening (2013, adapted)

• Providers discuss the risks and benefits of lung cancer screening in patients who meet the NLST inclusion criteria who are in relatively good health.
• Against screening patients who do not meet inclusion criteria or who meet the exclusion criteria
• Those opting for screening should get annual screenings until age 74, preferably at a site with experience in this method

Design

• Multicenter, parallel-group, randomized, controlled trial
• N=53,454 adults at high risk for lung cancer
  • Low-dose CT scan (n=26,722)
  • CXR (n=26,732)
• Setting: 33 centers in the US
• Enrollment: 2002-2004
• Analysis: Intention-to-treat
• Follow-up: median 6.5 years

Population

Inclusion Criteria

• Age 55-74 years
• ≥30 pack-year smoking history
• Quit smoking ≤15 years prior

Exclusion Criteria

• Lung cancer
• Chest CT in prior 18 months
• Hemoptysis
• Unexplained weight loss of ≥15 lbs in prior year

Baseline Characteristics

Derived from the CT group, which was similar to the CXR group. Notably, the participants were generally younger, had higher education levels, and were more likely to be former smokers compared to a 2002-2004 US Census tobacco survey.

• Age:
  • <55 years: <0.1%
  • 55-59 years: 42.8%
  • 60-64 years: 30.6%
  • 65-69 years: 17.8%
  • 70-74 years: 8.8%
  • ≥75 years: <0.1%
• Females: 41.0%
• Race:
  • White: 90.9%
  • Black: 4.5%
  • Asian: 2.1%
  • American Indian or Alaskan Native: 0.3%
  • Native Hawaiian or Pacific Islander: 0.3%
  • ≥2 race/ethnicities: 1.2%
  • Hispanic/Latino: 1.8%
• Smoking status:
  • Current: 48.1%
  • Former: 51.9%

Interventions

• Baseline demographic and smoking habit questionnaire +/- blood, sputum, and urine samples
• Yearly screenings for three years (T0, T1, and T2), no subsequent screenings for those diagnosed with lung cancer
  • Low-dose CT group - 1.5 mSv CT scan
  • Chest xray group - digital or screen-film techniques
• Analysis of images:
  • Single scan interpretation alone followed by comparison to previous scans
  • Positive ("suspicious for lung cancer") if non-calcified nodule ≥4mm (CT scan) or any non-calcified nodule (CXR)
    • Stable findings at T2 were reclassified as "minor abnormalities"
• Follow-up of masses: no mandated approach, guidelines were provided
• Patient follow-up:
  • Vital-status questionnaire once or twice yearly
  • National Death Index inquiries for those lost to follow-up
  • Medical records were followed regarding pathologic diagnosis and staging
    • Histologic classification: International Classification of Diseases for Oncology, 3rd Edition (ICD-O-3)
    • Staging: Cancer Staging Manual by the American Joint Committee on Cancer

Outcomes

Comparisons are CT vs. CXR.

Primary Outcomes

Lung cancer deaths

247 vs. 309 per 100,000 person-years (RR 0.80; 95% CI 0.73-0.93; P=0.004)

Secondary Outcomes

All-cause mortality

1877 vs. 2000 deaths (RR 93.3; 95% CI 1.2-13.6; P=0.02)

Lung cancer incidence

645 vs. 572 per 100,000 person-years (RR 1.13; 95% CI 1.03-1.23, P not given (pg 400, second column))

Additional Analyses

Non-lung-cancer deaths

Not significant

Positive result

T0 - 27.3% vs. 9.2%

T1 - 27.9% vs. 6.2%

T2 - 16.8% vs. 5.0%

Clinically significant abnormality, not concerning for cancer

T0 - 10.2% vs. 3.0%

T1 - 6.1% vs. 1.8%

T2 - 5.8% vs. 1.5%

Positive results resulting in lung cancer diagnosis

T0 - 3.8% vs. 5.7%

T1 - 2.4% vs. 4.4%

T3 - 5.2% vs. 5.5%

Stage of cancer with positive screening test

IA - 51.8% vs. 32.7%

IB - 11.2% vs. 14.9%

II A - 4.1% vs. 5.1%

II B - 3.1% vs. 4.0%

III A - 9.3% vs. 12.7%

III B - 7.7% vs. 9.8%

IV - 12.8% vs. 20.7%

Type of cancer with positive screening test

Bronchioalveolar carcinoma - 14.7% vs. 4.7%

Adenocarcinoma - 39.9% vs. 40.6%

Squamous-cell carcinoma - 21.1% vs. 25.4%

Large-cell carcinoma - 4.3% vs. 4.3%

Non-small-cell carcinoma "or other" - 34.1% vs. 10.1%

Carcinoid - 0.8% vs. 0.4%

Adverse Events

Complications following any invasive diagnostic interventions where lung cancer confirmed

Any complication - 28.4% vs. 23.3%

Worst complication is major complication - 11.6% vs. 8.6%

Worst complication is intermediate complication - 14.6% vs. 12.5%

Worst complication is minor complication - 2.2% vs. 2.2%

Death ≤60 days after most invasive diagnostic intervention - 1.5% vs. 3.9%

Complications following any invasive diagnostic interventions where lung cancer NOT confirmed

Any complication - 0.4% vs. 0.3%

Worst complication is major complication - 0.1% vs. 0.1%

Worst complication is intermediate complication - 0.3% vs. 0.2%

Worst complication is minor complication - <0.1% vs. 0.1%

Death ≤60 days after most invasive diagnostic intervention - 0.1% vs. 0.1%

Criticisms

• The authors identify:
  • Healthy volunteers may not be representative of the population as a whole
  • Modern scanners are more advanced than those in the trial, which may lead to increased detection of cancers or increased false-positives
  • The trial included institutions with radiology departments well-regarded for their abilities and may not be representative of the average radiology department of other institutions.
  • The reduction of death rate was only determined from three years of scanning, yearly scanning may provide more benefits.
• Letters to the editor identify:
  • Smaller benefit in per-protocol analysis than intention-to-treat analysis.
  • Psychosocial harm from the high number of false positives and invasive procedures unaccounted for.
  • Benefits of low-dose CT screening have not been clearly shown to outweigh the harms.
  • The authors undercalculated the rate of overdiagnosis by using the end of follow-up and not the screening period as the denominator.
  • Lack of cost-effectiveness calculations; one commentor calculates it as $725,000 for each lung cancer death.
• Did not publish methods of identifying cancer-specific deaths, which in the literature are notoriously difficult to categorize. A post hoc analysis of a subset of NLST patients suggested as high as 10% disagreement between providers in classifying the cause of death. If such a discrepancy truly existed within the NLST data, it could nullify the results of the trial.
• Generalizability is limited because of the strict inclusion/exclusion criteria, specifically the age for participant inclusion. For example, only 30% of lung cancers identified in the longitudinal Rotterdam Study satisfied the NLST's inclusion/exclusion criteria.
• The NLST criteria are not as sensitive as the modified PLCO criteria

Funding

National Cancer Institute.