Simple Blood Test Showing Promise in Lung Cancer Diagnosis

Jim Kling

November 07, 2014

AUSTIN, Texas — Researchers on a self-described "fishing expedition" have discovered something unexpected — that a panel of serum metabolites could predict the presence of non-small cell lung cancer.

The team compared blood from patients with a biopsy-confirmed diagnosis of non-small cell lung cancer with control subjects matched for age, sex, smoking history, lipid levels, and chronic obstructive pulmonary disease and diabetes status, explained by Peter Mazzone, MD, from the Cleveland Clinic Respiratory Institute, here at CHEST 2014.

"The metabolites we found make sense, knowing how cancer cells work, as to why there might be differences in these metabolites, but there's still quite a bit of work to do before this will have an impact on clinical care," Dr Mazzone told Medscape Medical News.

One potential pitfall of the study is that because so many comparisons were made, the statistical significance of about 5% would be due to chance. To counter that, the researchers used a q-value analysis, which sets much higher levels of significance to account for such artifacts.

The analysis involved 94 lung cancer patients and 188 matched control subjects. Mean age was 67 years, and 48% of the patients were women.

Of the cancer patients, 50 had adenocarcinoma and 44 had squamous cell carcinoma. The cancer of 44% was stage I, of 17% was stage II, and of 39% was stage III.

Cancer patients tended to be older than control subjects (68.7 vs 66.2 years; P = .013), but no other differences between the two groups were significant.

Each sample was divided and run separately, using ultra-performance liquid chromatography–mass spectrometry and gas chromatography–mass spectrometry. To identify metabolites, the researchers compared their analyzer outputs with a library of purified standards or recurrent unknown molecules.

They used random forest methodology to develop a prediction model.

The analysis identified 534 metabolites from eight metabolite superpathways and 73 subpathways. The concentrations of 149 metabolites were significantly different — 70 were lower and 79 were higher — between cancer patients and control subjects (q-value < 0.05).

Biomarkers to Rule Out False Positives

In patients with adenocarcinoma, the concentrations of 65 metabolites were significantly different from those in control subjects; in patients with squamous cell carcinoma, the concentrations of 50 metabolites were different.

The researchers built a model with 36 metabolites that, in a plot of sensitivity and specificity, had an area under the curve of 0.837. Sensitivity was 70.2% and specificity was 89.5% for distinguishing cancer patients from control subjects.

The models were generally more effective in identifying squamous cell carcinoma than adenocarcinoma, which came as a surprise, said Dr Mazzone.

"The next step will be selecting a model of the metabolites that are most discriminatory, and testing them in a brand new population of patients with and without lung cancer to see if those accuracies hold up," he explained. "For this to be clinically useful, the platform would then have to be simplified so it could be used in any clinic, as opposed to the research tools that we used."

 
Dr Mazzone and his colleagues cast a wide net, and they caught something. It's very interesting.
 

The required accuracy would vary depending on the purpose of the test. An upfront screen might require high sensitivity to minimize false negatives, and people with a positive result could be sent for imaging to rule out false positives. If the screen was used after suspicious nodules were seen on imaging, it might require greater specificity to distinguish cancerous from benign tissue.

"The challenge is in selecting the metabolites and thresholds properly, and informing clinicians how best they can be used, because false positives and false negatives can lead to bad decisions," said Dr Mazzone. A test could eventually help to profile cancers for prognoses, and to help determine which treatments might work best, he added.

The National Lung Cancer Screening Trial has shown a mortality benefit when former or current heavy smokers are screened with low-dose CT rather than traditional x-ray (N Engl J Med. 2011;365:395-409).

However, such methods lead to false-positive results, which present a dilemma, said Scott Oh, DO, assistant professor of pulmonary critical care at the UCLA Medical Center in Los Angeles, who attended the poster session.

Today's sensitive CT scanners find a lot of nodules that are benign, creating a very high false-positive rate and a dilemma for physicians who might not be eager to order expensive and invasive biopsies.

"Biomarkers are one of the ways to sort that out," Dr Oh told Medscape Medical News. Dr Mazzone and his colleagues "cast a wide net, and they caught something. It's very interesting."

The test will need to use simple technology that can be readily deployed in the clinic, and fewer biomarkers than this study used. "It will probably require a panel of biomarkers, but the fewer the better," said Dr Oh. "They'll also have to get down to a simple test that can be used in an everyday clinical scenario and not require a sophisticated lab and equipment."

Dr Mazzone and Dr Oh have disclosed no relevant financial relationships.

CHEST 2014: the American College of Chest Physicians Meeting: Abstract 1989452. Presented October 29, 2014.

Comments

3090D553-9492-4563-8681-AD288FA52ACE
Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.

processing....