Implications of a New Screening Test for Lung Cancer

Lung cancer is a devastating disease. It is one of those diseases, like ovarian cancer, that has often metastasized before it is detected. Certain forms of lung cancer can be particularly lethal, with mortalities of 50% or more within 6 months of diagnosis. That is why early detection is so critical.

Now we learn of a clever way of screening for lung cancer using… are you ready?… genetic profiling.

That’s right. Take 80 or so of the most likely genes from the DNA of the cells lining the patient’s airways and test them for the presence of certain gene sequences implicated in lung cancer to predict their likelihood of having lung cancer.

The authors from Boston University studied 129 patients undergoing bronchoscopy for the diagnosis of lung cancer that had a nearly 50% incidence of the disease in their study population (60 patients were diagnosed with lung cancer). Their test utilized microarray genechips made by Affymetrix for their analysis. The authors acknowledged an 80% sensitivity and 83% specificity with their test – far higher than other tests available to date, but had some important limitations:

In the setting of our study, where disease prevalence was 50%, a negative bronchoscopy and negative biomarker for lung cancer resulted in a 95% negative predictive value, potentially allowing these individuals to be followed nonaggressively with serial imaging studies. For individuals with a negative bronchoscopy and positive gene-expression signature, the positive predictive value was 70%; these individuals would probably require further invasive testing to confirm the presumptive lung cancer diagnosis. However, compared to bronchoscopy alone, the strong negative predictive value of the combined cytopathology and gene-expression biomarker test should substantially reduce the number of individuals requiring further invasive diagnostic testing.

The notion of a cancer-specific airway-wide injury suggests that cancer-specific alterations in gene expression that occur as a result of smoking might precede the development of lung cancer. If this is true, the lag between alterations in gene expression and the appearance of lung cancer could contribute to the biomarker's false-positive rate in our cross-sectional study. A longitudinal study will be needed to assess whether false-positive biomarker diagnoses represent smokers at higher risk for developing lung cancer. If this is the case, our biomarker might be useful as a screening tool for lung cancer among healthy smokers and may have the potential to identify high-risk smokers who would derive the most benefit from chemoprophylaxis.

So What Do These Data Mean?

Well, I’m not sure we’re ready to go the chemoprophylaxis or even the screening of smokers route just yet. To better understand the implications of a test with 80% sensitivity and 83% specificity, one must understand the statistical definitions of these two terms. For sensitivity, it means the probability that a test is positive in the population of people known to have the disease. Specificity, on the other hand, is the probability that a test is negative in the population known not to have the disease.

So how can these numbers be applied here? Doing some back-of-the-envelope calculations from the authors’ data, their test with its 80% sensitivity in 60 people known to have lung cancer means that 48 had a positive test that appropriately detected the cancer while 12 had cancer that the test failed to detect. Likewise, if 69 patients did not have cancer and the test had an 83% sensitivity, then 57 patients who did not have cancer tested appropriately negative for the disease, but 12 tested positive for the disease, even though they did not have cancer.

Now, in this population with a strikingly high incidence of cancer, the reliability of this test is reasonable, but in the population at large (i.e., all smokers) where the prevalence of disease if likely to be much lower, there will be far too many false positive tests (diagnosing cancer in people actually free of disease) and importantly, still a significant false negative rate (thinking a person is free from cancer based on a negative test, when, in fact, they have a cancer).

So there’s still some work to be done before such a test can be applied to the general population. But the study authors should be commended on their clever use of micro-array DNA genomic phenotyping to improve the detection of lung cancer during bronchoscopy over current cytologic (cell analysis) techniques.