DR. HOFFMAN: My job here is to talk to you a bit about what is next on the CT side of imaging, paying particular attention to the use of contrast CT.

As we go through the first part of the talk, I'll just put some of what we are talking about with contrast CT in the context of what portion of these lesions are we looking at with contrast. For this I borrowed some slides from the group at the Mayo Clinic, from their initial lung cancer screening study, and then later the multicenter trial, looking at CT contrast enhancement.

This is the Mayo approach. You find a nodule of less than 3 millimeters. At six months you do a thin section CT, and then based on that, you go from there.
If the nodule is 4-7 millimeters in size, you ask the patient to come back in three months to recheck it. If the nodule is 8-20 millimeters in size, you do thin section CT immediately. Based on the visual characteristics of that, you either go straight a nodule enhancement study with CT, and based on that, you can go onto PET. Otherwise, you continue to follow.

In a prevalence study looking at 1,500 patients, 782 of these patients, 51% were found to have nodules. And of those 782 patients, there was 1,300 nodules. This shows the number of nodules in a given set of patients, and the size distribution of nodules.

I put some of these up here to show what we are talking about in the current protocols for contrast CT. We are talking somewhere on the order of about 11% of these nodules in the Mayo protocol are felt to be appropriate for contrast CT. As we'll talk about a little later, my suspicion is that this protocol has developed out of the days of single slice, slightly thicker slice spiral CT. And with newer, thinner, easier to acquire CT slices, that we may want to move up this ladder in terms of contrast and characterization early.

On a second scan baseline plus first annual scan the number of nodules detected are creeping up here. So the number of studies that potentially, under the current protocol, could go onto contrast CT, and in particular, in the current protocols that go on to a much more expensive PET study is getting rather large.

Of those 1,500 studies, 25 were cancer, to put things in perspective. And 7 that went on to resection were benign nodules. So clearly, there is a role to improve our ability to characterize these nodules, and to characterize in as simple a way as possible. The premise on the CT side is that it's much less expensive than PET if we can make it as sensitive and specific.

So the conclusions of that first Mayo study, among the 8 millimeter nodules, 22 of them were eventually proven to be malignant. Only 1 of 2,264, 0.04% of the nodules less than 7 millimeters detected were eventually proven to be malignant. And therefore, there needs to be a clearer, inexpensive, simple way of following these small nodules.

There were a number of papers looking at nodule enhancement. FDG PET, as people talk about it being a metabolic scan, as I talked to PET people around the country, most people are in agreement, or at least uncertain as to whether or not we are really looking at metabolism or whether we are just looking at blood flow, PET's version of blood flow. You heard earlier that there certainly are very exciting, new PET labels that are clearly not just measuring blood flow. But DeJuan and colleagues, sensitivity, specificity, 95%-80%, sensitivity, specificity in PETs, 100%; 89%, 93%, 88%, 95%, 87%, which is fairly good.

In CT contrast studies, there are smaller numbers here, 32 patients. There isn't a sensitivity and specificity specified here. Swenson and colleagues at Mayo, 100% sensitivity, 76% specificity, a little lower in that study than the PET.

Here are 25 patients, sensitivity 100%, specificity 74%. Some of those are a little lower than the PET. This is a multicenter CT contrast trial, here are the participating centers, the frequency, and the percents.

The protocol here again was using an older spiral scanning technique, 3 millimeter collimation, single slice spiral, 120 KV, 280 MA, picture one-to-one covering 15 millimeters of Z axis span. The protocol was to inject contrast and scan every minute post-contrast for four minutes. And they gave 2 ml per second, 300 milligrams of iodine per ml, 420 milligrams of iodine per kilogram body weight.

The number of nodules studied with this was 356. It was felt that a threshold was set to identify malignant versus benign of 15 Houndsfield units. Anything below 15 Houndsfield units was considered benign. The question being asked was how well can they characterize and identify benign tumors, keeping them out of the move towards more expensive studies? Sensitivity was 98%, specificity was 58%, and the accuracy was 77%. The prevalence, let me just step through quickly to some of the punchline here.

With the advent of multi-slice scanners, we can now follow a contrast agent, not at one minute intervals with slow contrast drips. But we can follow the contrast as a first pass. And we can do this in as small as half millimeter voxels at a time.

We have shown with CT scanning and deconvolution methods we can get sensitivity down to the point where we can give you the timing of a red blood cell through the capillary bed of the lungs in an individual voxel. So that the contrast CT studies that have been done to date characterizing lung nodules, looking at 58% sensitivity, looking at one minute intervals, et cetera, clearly have vast room for improvement in characterizing these nodules.

This is a plug here just to discuss the fact that we are characterizing the lung, looking for nodules in patients that are at the high risk. Most of these patients are at high risk because of their smoking. Because of their smoking, most of these patients probably have early emphysema as well. And I just thought I'd take the platform for a minute to plug that maybe we should be doing a comprehensive assessment of the lung, not just looking for nodules. The emphysema that can be detected in these patients is probably a much greater health problem, and much greater potential cost to society than even the cancers.

We have a tissue characterization protocol where we have trained the computer, based on what looks normal in the COPD subject, to train the computer to recognize that as early abnormal. If we scan a set of normals and a set of smokers, both with normal pulmonary function tests and whose lungs to the radiologists look normal; with 80% accuracy, we can tell you who the smokers are, and who the nonsmokers are. And these lungs are not normal, not just from cancer, but also because of emphysema.

And then just to finish off on this theme of tissue characterization, I'll show a few slides from UCLA, where they have followed a similar protocol to the multi-center trial, scanning at one minute intervals. But rather than just looking at the overall brightness change in the tumor, they looked at the pattern of brightness change.
And here you can see these very intriguing early studies of different patterns of contrast enhancement, not only low contrast enhancement in the benign tumors. So that that 58% specificity, I think that based on looking at patterns of enhancement, looking at first pass studies, looking at thin slices using multi-slice, soon the scanners will have 16 slices at a time, 0.75 millimeter slice thicknesses, that we can do much better with CT than we are doing at the moment.

With the new technology we will probably have to completely revisit, I would say, the comparisons of sensitivity and specificity of contrast CT to characterize these tumors. We will also need to rethink whether or not we should be doing contrast with these new techniques on much smaller lesions.
Thanks.

TOP