A recent research project commissioned for the Lung Cancer Alliance and several other interested organizations uses actuarial techniques to calculate the mortality difference between people who are diagnosed with early-stage lung cancer and those who are not diagnosed until lung cancer has reached its later, fatal stages. We recently sat down to discuss the implications of this paper with the authors, actuaries Bruce Pyenson, Sara Goldberg, and Dale Hagstrom, and cancer researcher Dr. James Mulshine, as well as with Laurie Fenton and Sheila Ross of the Lung Cancer Alliance. This interview puts the new research in a broader context—as part of the larger effort to reduce the mortality associated with a disease that kills more than 160,000 people every year.
Q: How is lung cancer different from other cancers?
Jim Mulshine: Lung cancer is an exemplar of many of the lethal cancers. In fact it is the most lethal cancer, accounting for almost a third of total cancer mortality. But lung cancer, esophageal cancer, head and neck cancer, ovarian cancer—there are a variety of these diseases that we really haven't positively impacted at all. The shared feature of these cancers is that we find them relatively late in their natural histories. So at the time that the average patient with one of those lethal cancers is diagnosed, the cancer has spread to other parts of the body, at which point we cannot cure those patients. The frustration with lung cancer is that there is an obvious opportunity if we can fight the disease in a systematic fashion, prior to the spread of the cancer to its lethal phase. Early detection, which has been the key to turning around mortality outcomes with cervical cancer, breast cancer, and other cancers, is an obvious strategy for lung cancer. But our effort to systematically implement a strategy of earlier detection for lung cancer is something that the general public has a lot of suspicions and concerns about.
There are three reasons for the suspicion and concern. First, we've been talking about this for a long time, and it's not obvious that we've made much progress against cancer, which introduces a certain amount of skepticism. Second, inherent in an early detection strategy for a cancer is a pattern of care that we're not comfortable with. We're comfortable with people getting sick and going to see their doctor. It's counterintuitive to many in our population that the time to find a disease is when you're feeling perfectly well. Third, our medical society is really not set up to do this kind of preemptive strategy. We don't have the public health infrastructure to do these things in a highly standardized, quality-controlled fashion. Regionalization of care, which is a characteristic of many societies, is not a characteristic of our society and it comes into clear focus in the area of cancer preventive services.
Sheila Ross: Tying in to Jim's point about early detection: One of the challenges is that lung cancer is a slow-developing cancer, so it's going to take 20, 30 years to develop, because you have a long lead time—which also makes it ideal for screening, ironically enough.
Q: How does the potential of lung cancer screening compare to other kinds of cancer screening?
Jim Mulshine: From an a priori perspective, lung cancer is a leading cause of cancer-related death in our society, and the average person diagnosed with lung cancer has a 15% five-year survival expectation. Prostate cancer is a common cancer, but much less lethal. The average patient diagnosed with prostate cancer has a better than 99% five-year survival. So in a sense, prostate cancer is a much more indolent disease process than lung cancer and the upside opportunity of finding lung cancer earlier is much more significant than for prostate cancer. The main test for prostate cancer is a blood test that takes advantage of the buildup of a protein in the blood. That elevation of the protein could be related to benign prostate tissue growth, which does occur fairly frequently in older men, or it could be related to aggressive cancer. And so the signal versus noise or the ability to distinguish lethal cancer events from other circumstances with the PSA testing has always been of concern to a lot of people. Meanwhile, spiral CT screening, which is the most advanced approach to early detection of lung cancer, has performance characteristics that are much more objectively favorable.
Sara Goldberg: By the time a typical patient is showing lung cancer symptoms (which can often be misdiagnosed as pneumonia), they are frequently on to Stage IIIA, IIIB or IV—all late stage. Lung cancer diagnosed at this stage translates to very slim odds of five-year survival—less than 5% for Stage IV. Those odds are much worse than regional or even distant stages of other cancers like breast and prostate. Our study estimates a very real mortality difference with early detection: five-year survivorship of the same group of lung cancer patients increases to over 50% in our baseline model, even after accounting for "lead" time.
Q: Presumably if screening happens, it's not screening for everybody but rather for certain people. Who would you characterize as at high risk for lung cancer?
Jim Mulshine: Right now, the tools for identifying screening candidates are smoking exposure and age. While they're not perfect, they define a fairly high-risk population. Using the traditional tools, the average pack years (years of smoking one pack a day) of people diagnosed in the studies have been in excess of 40. And the reason for that is that 90% of people who smoke will start by age 20, so that the number of people that are 60 years old that have smoked for ten years is infinitesimal. Very few adults start smoking at age 52.
Age and smoking status have been the traditional tools. But over the last three, four years, there's been an explosion of new research done on this—MD Anderson, Liverpool Lung Project, David Christiani's group at Harvard—in which new risk tools have been defined using a combination of epidemiological tools and molecular diagnostics. The ability to define high-risk populations has improved considerably.
Sheila Ross: Think about who gets lung cancer: 35% of those being diagnosed with lung cancer are current smokers, 50% are former smokers, most having quit decades ago, and another 15% have never really smoked. In fact, if you just look at the frequency of lung cancer in never smokers, that population is still the sixth biggest cancer killer in the United States.
It used to be that the majority of lung cancers diagnosed were squamous cell lung cancer, which seems to have a very definite association with smoking. But over the past 20 years now, we have seen adenocarcinomas and other different cellular types emerge, and adenocarcinomas have now become the most frequently diagnosed lung cancer. Squamous cell used to be the majority, and was at 60% or 70%. Now it's down to about 20%. So it's not just smokers who are at risk.
Q: So how do we clear both the cultural and technological hurdles and bring about more effective lung cancer screening?
Sheila Ross: We understand from people who were there at the time, after the Surgeon General's report came out in 1967 demonstrating the link between smoking and lung cancer, that public health policy people faced some difficult question. They thought that the quickest and most cost effective way to handle the findings would be to demonize smoking. They couldn't take on the tobacco companies head on—they were too powerful—but if they demonized smoking, they could force a social change that would then force people to stop smoking or not take it up. The strategy worked. In the '60s, ashtrays were everywhere, and even in Congress there were ashtrays on the back of every chair in hearing rooms. It was just ubiquitous. Certainly that has changed. But in the midst of all this, the smoker himself was demonized, and that's an unfortunate part of it that has contributed to current stigmas.
Laurie Fenton: Consider that smoking is implicated in many, many cancers and diseases—yet lung cancer is the only disease "blamed" on the patient. This is tragic and the consequence has been little compassion and support for people who do not "deserve" this lethal disease. No one deserves lung cancer—no one.
As Sheila stated earlier—when you consider that today 65% of those diagnosed with lung cancer will be former and non-smokers—a tobacco prevention and control strategy alone will not address the needs of this majority. We must, in addition to these efforts, ensure that early detection and better treatment options are available and sufficiently resourced to complement prevention and control strategies. Only then will we have a comprehensive plan that will reduce lung cancer's mortality. Early detection research plays a central role, as that is what has contributed to survival increases in breast, colon, prostate, and cervical cancers, for example. Imaging is the current option available today—while we continue to push for other biomarker breakthroughs of tomorrow.
Q: What are the treatment options for lung cancer?
Sheila Ross: The best treatment option is surgery. There's a long-term observational study that is still going on, the International Early Lung Cancer Action Program (I-ELCAP), and they found that a screening of high-risk populations could diagnose lung cancer at Stage I, and 85% of the time, and that of those diagnosed in Stage I and treated immediately with surgery, they had a 10-year life expectancy of 92%. That’s a complete reversal of the statistics we have now.
That surgery is getting easier, too. They are doing wedge resections instead of taking out whole lobes. Video-assisted surgical techniques have greatly reduced the size of the incision and the time and cost of recovery.
Q: How does Milliman's analysis of lung cancer screening add to the discussion?
Laurie Fenton: This study is so important because it provides an additional set of data points to validate the premise that finding lung cancer early can save lives. Bruce, Sara, Dale, and Jim should be commended for their efforts as this study proves that when detected early, lives can be saved.
This actuarial analysis helps answer questions of lead time bias, length bias, and over-diagnosis—which are almost impossible to answer in a randomized control setting. Sadly, it is exactly these biases that have been used as justification against lung cancer screening. With the new insight, these findings should be used to encourage the development of uniform standards and guidelines for lung cancer screening.
Bruce Pyenson: There's a heated set of arguments on potential biases for lung cancer screening. There are also some pretty standard actuarial tools that have been used for decades by life actuaries that could really model those biases very well. These biases raise questions about the connection between early detection and mortality improvement. For example, are people living longer after diagnosis just because you found it earlier, and you're starting the clock earlier because you found the cancer earlier? Our study shed light on this bias by showing that when someone gets early stage lung cancer and gets treated, their survival is so much better—it's hard to explain away.
Jim Mulshine:In the United States, 160,000 people die each year of lung cancer, and most people think that's just the way it is, and there's nothing we can do about it. This paper challenges that conventional wisdom and gives us a chance to calibrate the opportunity. Each year, more than 200,000 new people are diagnosed with the disease. We have tools that aren't particularly expensive or unavailable that could be applied in a very carefully defined way and potentially avert some of that mortality.
Q: Does the actuarial analysis of lung cancer mortality tell us something that randomized clinical trials cannot?
Jim Mulshine: Actuarial approaches are used in so many walks of life, and here you come into this area where a lot of people say that lung cancer screening doesn't work, because there's no randomized trial that shows that it's effective. And the reality is that the technology to do this is relatively recent, and there's been no randomized control trial that's been completed. Now you take these actuarial techniques with very mature data sets, with methods that are well validated in other areas, and you apply them in an area where there's a vacuum of data. The actuarial approach gives you a tool to approximate potential benefit.
Unfortunately, it is becoming more apparent that assessing the value of a screening approach is exceedingly complex. We have no single tool that is satisfactory. As we are seeing with the randomized evaluation of spiral CT screening, this approach is requiring about 15 years to complete. In the initial discussions of this trial, the goal was to complete the trial in 10 years. In the time to complete the randomized spiral CT screening trial, well over two million additional people in the United States will have died of lung cancer. When the Health Insurance Plan (HIP) study evaluating the benefit of breast cancer screening was performed, that trial took 26 years, and involved more than 60,000 women. However, mammographic technique did not rapidly evolve. If the current randomized CT study does take15 years to report, then several generations of CT technology will have evolved. The CT equipment used in the screening trial will have different performance capability from the CT equipment deployed in 2015, so there will be a question of relevance. Having additional analysis approaches to evaluate the benefit of cancer screening approaches is an important innovation.
Dale Hagstrom: A lot of controlled experiments look at two sets of people—those receiving the treatment and then a control group. Mortality studies are a little different because when you're talking about a death, that's a statistically volatile event. It is hard to find a specific control group and measure them because they themselves are subject to their own statistical volatility. You lose a lot of the information when you try to pair up a treated group and a control group one by the one, because the mortality rate is itself rare in any one year for the normal person. By using a whole mortality table, however, your control group in effect is a huge population over time and still allows the comparison to be matched on age, gender, race, and smoking habit. So you’re comparing the group receiving treatment against a group that won't generate noise by itself. This particular approach has that advantage—of using a mortality table to get around the control group problem.
Q: Are there aspects of the delivery of lung cancer screening that need to be improved if we are to satisfy this mortality-improving potential?
Laurie Fenton: Consider breast cancer and mammography. Today we have Congressionally mandated uniform standards and guidelines that govern the practice of breast cancer screening—implemented to reduce harm and risk for women undergoing the procedure. People are being screened for lung cancer today—thus we need to emulate the same approach taken by the breast cancer community and enact uniform standards and guidelines to best manage the practice of lung cancer screening. This has clearly benefited the breast cancer community—why should we have a different standard for the lung cancer community?
Jim Mulshine: One incredibly important issue is how our society organizes preventive services. One possibility is the creation of a national center of Jiffy Lube-type dedicated service centers to do multi-point preventive services. If we could have a matrix in which people could get efficient, high-quality care with informatics linking all these preventive services, it would have a profound impact on the health of our nation.
Q: Are there any mechanisms in place for building out this kind of lung cancer screening apparatus?
Jim Mulshine: I-ELCAP, which has already been mentioned, has been going since 1999 and has more than 60 screening centers. They have over 50,000 people in follow-up. It’s a continuously evolving research platform, in which they link all their trial participation through a confederated structure. Each site owns its own data, but the aggregate results are owned by the association.
Bruce Pyenson: I-ELCAP employs a detailed level of protocol and seriousness in the centralized collection of data. Unfortunately the opportunity to transform knowledge through huge databases is controversial. For example, there are observed differences that some hospitals in Oregon are keeping people about half as long as hospitals in New York, and the outcomes seem the same. But that's observational—it's not a randomized control trial—so some question the conclusion that New York hospitals are inefficient.
Q: What is the next step in the move toward better lung cancer screening and intervention?
Jim Mulshine: The paralyzing aspect of screening for lung cancer is that every other adult in our society is either a current or a former smoker. So how do you handle 100 million people at risk for this terrible disease and not break the bank on our already teetering healthcare system? And I think that's where future analysis will be very valuable. It may be that if you move beyond just direct cost, and you look at total cost, and total benefits, that there is a clear economic benefit in a 55-year-old not dying acutely of lung cancer over a very catastrophic six-month period of time.
These imaging technologies are finding cancers that are about one and a half centimeters. And the screening cancer cases that are discovered in annual follow-up, just like with breast cancer, are very small—they're eight-tenths of a millimeter. Ten years ago, we virtually had no eight-tenths of a millimeter cases of lung cancer detected. Through the refinement of screening approaches they're finding smaller disease. Meanwhile the surgery required to remove that completely and effectively is becoming less and less intrusive. Minimally invasive approaches are being refined and generalized across the country, so that the cost of hospitalization and related costs related to lung cancer care are falling. We've got to look at the economic impact in a way that allows us to make smarter decisions about where we invest healthcare resources, If we determine all of the life cycle costs and benefits of systematic early lung-cancer detection, with continuous process improvement, there is already some data suggesting that the cost of providing very efficient lung cancer screening services may be much more manageable than people expect.