News

Predictive models revealed lung nodules’ cancer risks


 

FROM THE NEW ENGLAND JOURNAL OF MEDICINE

New statistical models can be used to predict with "excellent" accuracy which lung nodules found on screening low-dose computed tomography scans of high-risk patients are likely to be cancerous, according to a report published online Sept. 4 in the New England Journal of Medicine.

"Our models are coupled with risk calculators, which make possible the rapid and easy calculation of lung-cancer risk given the characteristics of the person and the nodules," said Dr. Annette McWilliams of Vancouver General Hospital and the British Columbia Cancer Agency, and her associates (N. Engl. J. Med. 2013;369:910-9).

Several previous studies have reported that more than 20% of people who participate in low-dose computed tomography (LDCT) screening programs are found to have at least one lung nodule that requires further investigation. Yet only a small percentage of these nodules turn out to be malignant.

In one large screening trial, 25% of the surgical procedures that were prompted by an initial LDCT scan were performed on nodules that proved to be benign. That means that many patients were needlessly exposed to invasive diagnostic procedures, Dr. McWilliams and her colleagues said.

They developed models that would accurately predict the likelihood that such nodules are malignant, so as to guide clinical decision making and thereby reduce the risks and the costs of morbidity and mortality in LDCT screening programs.

The first step was to collect a large amount of detailed data regarding nodules identified on LDCT that were either cancerous when detected or later developed into cancers. To do that, the researchers analyzed data from the multicenter Pan-Canadian Early Detection of Lung Cancer Study, a population-based sample of current and former smokers aged 50-75 years who had no history of lung cancer but had a 3-year risk of at least 2% of developing the disease.

A total of 1,871 of the 2,537 participants in that study (74%) were found to have 7,008 lung nodules on initial scanning. They were followed for a median of 3.1 years using repeat LDCT at 3- to 12-month intervals, depending on the size of the largest nodule.

Lung cancer was diagnosed on the basis of histopathologic assessment of resection specimens or cytopathologic examination of needle-aspiration biopsy samples. A total of 102 nodules (5.5%) were found to be malignant.

Tumor characteristics that strongly correlated with malignancy were larger nodule size, partially solid type (as opposed to nonsolid, fully solid, or perifissural types), location of the nodule in the upper lung lobes, a small number of nodules (rather than numerous nodules), and the presence of spiculation.

Patient characteristics that strongly correlated with malignancy were female sex, older age, family history of lung cancer, and emphysema.

Dr. McWilliams and her associates used these traits to develop prediction models. They then tested the models in a validation cohort: 1,090 current and former smokers aged 50-74 years who underwent LDCT as part of several chemoprevention trials conducted by the British Columbia Cancer Agency over a 10-year period.

Those study subjects were found to have 5,021 lung nodules. Forty of those patients were found to have 42 lung cancers during a median follow-up of 8.6 years.

"Our models showed excellent predictive accuracy," clearly differentiating study subjects who had or developed lung cancer from those who did not. "The models performed well even when applied to nodules 10 mm or smaller, which are the most clinically challenging and most numerous nodules," the investigators noted.

None of the 571 perifissural nodules found on LDCT in those two cohorts proved to be malignant. "When the data from the two studies were pooled, the probability of lung cancer in perifissural nodules was 0," Dr. McWilliams and her associates said.

Such nodules "probably do not require longitudinal follow-up with CT," they added.

Their models are not intended for use in low-risk people for whom LDCT is not currently recommended. The models also "do not apply to persons with hilar or mediastinal lymphadenopathy, for whom further investigations are warranted irrespective of the nodule size," the researchers added.

The Terry Fox Research Institute, the Canadian Partnership Against Cancer, the U.S. Public Health Service, and the National Cancer Institute supported the study. Dr. McWilliams reported no financial conflicts of interest; her associates reported ties to CareFusion, Eli Lilly, GlaxoSmithKline, Med Biogene, Olympus, Pfizer, Roche, Siemens, and Toshiba.

Next Article: