On March 24, 2009, Dr. West wrote,

Lung cancer screening is one of my least favorite topics to discuss because it’s probably one of the biggest areas where there is a gulf between the medical establishment’s party line and the expectations of many patients and advocates. I tackled a discussion of screening a few years ago that included the anticipated benefits as well as the challenges with LC screening (nowadays really focusing on low dose, spiral CT). That was probably about the most frustrating topic I’ve pursued, initially heralded after my post on the arguments in favor of screening, but feeling like I was being vilified as a kitten torturer in the responses I received after my post about the thorny issues with it.

Dr. West’s last post is relevant to the conversation that member cards7up and I have been having on the discussion thread from my piece on November 30, 2009. The gist of the conversation is about whether to screen, and if so, how to do it.

On the “if” question, I consider my father’s desire for pan-scans to look for occult cancer (he quit smoking about 20 years ago). Every now and again, he hears on the radio ads for commercial enterprises offering pan-scans for a fee to screen for cancer. Even though he can afford it, I don’t let him go—I feel that he’s more likely to be harmed by the test than helped. Dr. West addressed this issue in his discussion about PSA and prostate cancer:

But last week, the story emerged that PSA screening for prostate cancer appears to provide somewhere between little and no survival benefit for men. One large trial suggested a very small benefit, in which nearly 50 patients would need to be treated to save the life of one from prostate cancer, and another trial didn’t show a survival benefit. After lung cancer, prostate cancer is the next most common focus of my practice, and I can assure you that I find this very easy to believe. I see many men who are exceptionally likely to do well, begging the question of whether they underwent surgery or radiation or some other alternative for no benefit, but who have had a permanent compromise of their sexual function, urinary continence, and other issues important to their quality of life (and I probably see only a small minority of the patients who are exceptionally likely to do well and never even see an oncologist). I also see men with prostate cancer who undergo aggressive treatments for prostate cancer but have a virulent enough cancer that even screening efforts and timely, aggressive interventions can’t save them from their cancer. In between, the numbers suggest one in 50 men may be saved from a truly life-threatening cancer.

Lung cancer is not prostate cancer, but certain principles are shared. The real goal of screening of smokers, as in any cancer-care decision, must always be to improve duration and quality of life; everything else is a surrogate measure. If screening should cause more suffering and deaths from complications of unnecessary biopsies and surgeries than it saves by catching early stage cancer, then it should not be done. If it can save lives and reduce suffering, it should be done. We do not know this answer yet nor do we know the best way to screen.

The latter is the question that cards7up raised. It’s a very important question that can influence the question of whether we will provide a net benefit by screening. A good screening test for lung cancer would find cancer in everybody who has it, be negative in everybody who does not have it, and pose no risk to the patient. The closer that a test gets to these characteristics, the more likely it is to be of net benefit (and by enough to justify the cost to those in government and insurance who will ultimately be faced with financial cost/benefit decisions). Most of the screening trials use low-dose imaging (to minimize the radiation) and no contrast (to avoid the risk of allergic reactions to contrast die and the risk of damage to the kidneys). The simple answer to the question originally posed is that for those who choose to get screening before we know if it helps or not, I favor a low-dose, non-contrast scan to help minimize that chance of harm; there is no proof that this the “right” answer. Also, those who choose to get screened should be aware that most insurance will not cover it, so there may be substantial financial toxicity.

With these questions in mind, I read with interest this week’s New England Journal of Medicine (by happy coincidence, just an hour before seeing cards7up’s post!) We have seen preliminary data from the NELSON trial before. Briefly, this trial randomized high risk patients to low-dose CT or not, with 7557 getting screening. Screened patients got CTs at 1 year, 2 years, and 4 years. The trial seeks to evaluate whether the screening strategy can reduce mortality from lung cancer by at least 25% at ten years. This is a good choice of endpoint—a 25% reduction in death would be extremely important and would argue strongly in favor of screening. The data are not yet mature to ten years, but we do have some preliminary data on the operating characteristics of the test that can inform our conversation. First, 63.9% of cancers found were at stage I—this is promising, as the cure rate for these cancers is reasonably high and they are typically asymptomatic at this stage (thus not detected until they are more advanced).

This paper does not give us the answer to whether and who we should screen and it doesn’t try to. What’s cool about this paper is a technique that the investigators used when dealing with indeterminate non-calcified nodules that may or may not be cancer. You’ll recall that Dr. West raised the concern about harming patients with screening and I seconded it. In particular, I worry about harm, in terms of anxiety and unnecessary procedures (with complications) for people who get a positive scan, but don’t really have cancer. If we can improve the test and lower this rate, then we could help ameliorate that concern and potentially improve the value of testing. In particular, small non-calcified nodules that show up on scanning are often not cancer. The investigators used a computer-based algorithm based on volume of a nodule and its rate of change over subsequent scans to improve evaluation of these nodules.

So how good were the results? In the first scan, the chance of finding a cancer, if present (sensitivity) was 94.6%. The chance of ruling out disease, if absent (specificity) was 98.3%. Sensitivity and specificity can be restated as the chance of a result (once you have it) being true. These values are called positive and negative predictive value and they can change dramatically from the sensitivity and specificity based on the prevalence of the disease (how likely a person is to have lung cancer regardless of testing). The chance of a positive test result really representing lung cancer (positive predictive value) was 35.7% and the chance of a negative test result really representing absence of lung cancer (negative predictive value) was over 99%. So, at first screening, a negative test result was very reassuring, but a positive result needed further evaluation, with a high chance of not really meaning cancer. In the second round (year two), sensitivity was 96.4%, specificity 99%, positive predictive value 42.2%, and negative predictive value 99.9%.

So, how many cancer deaths were prevented and how many people were harmed by unnecessary procedures? Although the authors provided some information on followup in the online appendix, they did not describe objective complication rates, and anxiety is hard to quantify. No matter whether you believe in lung cancer screening passionately or have equipoise (a belief that either approach is appropriate until we get more information) while the screening trials data mature, we can all agree that better screening is a good thing. When will we find out the “hard” endpoints from NELSON? Round two screening was completed in April, 2008, so round three should complete roughly two years later, with ten year data available by 2015 or 2016. However, given how aggressive lung cancer can be, the authors may have meaningful data on hard endpoints sooner than that if they perform an interim analysis. I for one, would like to see more lung cancer cured at early stages. Having finally met Dr. West in person at a lung meeting in DC a few days ago, I feel fairly confident that he feels the same way (and does not torture kittens).