For background on this post see a recent New York Times article, and also a special issue of the periodical, The Cancer Letter:
Lung cancer has always been a problem for physicians advocating screening and early detection. For a long time no one was able to show that by the time a lung nodule gets big enough to be seen on an x-ray or a CT scan, that there was anything you could do to save the patient's life or significantly alter the course of the disease. Therefore the various groups that recommend screening tests for cancer did not, as a rule, recommend screening for lung cancer.
At least two powerful groups did not like this. One was the tobacco industry. They wanted to be able to reassure smokers that they could reduce the risk of dying of lung cancer while still puffing away with abandon (and profit to the companies). The other was the big companies like General Electric that make CT scanners and that want to push the use of these expensive devices.
These two groups found some willing scientific collaborators based at Weill Cornell Medical College in New York. Money flowed into the group from the tobacco industry and GE, and they supplemented that cash with some hefty Federal grants. The group made headlines by publishing a paper in 2006 claiming that they could prevent as many as 80 percent of lung cancer deaths with their screening protocol.
Those who knew about the basic epidemiology and biology of lung cancer thought these data too good to be true from the get-go. Their suspicions were deepened when it was revealed in 2008 that the chief scientists had these conflicted relationships with powerful industry groups with a stake in the outcome of the study.
But it gets worse, as it turns out. The sources listed above are based on a review that Weill Cornell conducted, and then apparently never revealed or acted upon. The review did, however, apparently result in the two chief scientists relocating to other institutions.
The major publicity in the past week or so has focused on one easy-to-understand lapse. It turns out that the review team could not locate 90% of the subject consent forms, meaning that in a very basic way, this research failed to meet the criteria for ethical research. The chief investigators made the lame protest that the study was conducted at 37 institutions, and that each center was responsible for maintaining its own set of consent forms. But the fact remains that in publications, the chief authors assured readers that all subjects had given formal consent, and that it now turns out they had no clue as to whether or not consent was obtained.
But a closer reading of The Cancer Letter reveals other basic flaws. Apparently the study protocol failed to assure that uniform standards were followed at all 37 participating centers. Also, no effect size calculation was ever conducted to determine how many subjects needed to be enrolled to prove or disprove the study hypothesis; so essentially this is a never ending study. (And indeed, since the review was sent to Weill Cornell, a further 10,000 subjects have been recruited, which is completely unethical if the basic protocol has been questioned.) In other words, cutting to the chase, as science, this study was a piece of junk--and journals who published data from the study need to seriously consider retracting the articles.
Despite being a piece of junk, the research may have gotten something right. A later study that appears to have been sound, that was Federally funded, concluded that lung cancer screening could be of value in heavy smokers. But while agreeing with the overall direction of the Weill Cornell findings, the Federal study found an effect size much smaller than the 80% claim. (For the reasons I started off with, skeptics might worry that even this Federal study might not be borne out in future research.)
I dare not say if there is any general lesson here about conflicts of interest in research funding. Certainly you can have conflicts of interest and still do research that is scientifically valid and that follows the proper rules of human subjects protection. But in terms of research not passing the sniff test, and then later turning out to smell worse and worse the more you examine it, this group of studies seems to have set the curve.