How would you best care for the President? Why Biostatistics should matter to every physician
I made an astonishing discovery recently. There’s a screening program that could have reduced President Barack Obama’s risk of dying of lung cancer by 20%, or even 33%, while he was in office. Were his physicians careless, or does a “33% reduction in risk” not tell the whole story of risks and benefits to lung cancer screening? How might more selective care actually have been better care for the President?
A known former smoker, President Obama is at an increased risk for developing lung cancer, which causes over 1/4 of all cancer deaths in the United States (more than breast cancer, colon cancer, prostate cancer, bladder cancer, and lymphoma combined) and which has a 5-year survival rate of only 17.7%. In 2013, President Obama signed into law the “first legislation requiring comprehensive plans of research action for high mortality cancers, with lung and pancreatic cancers given priority status.” More recently, in 2016 President Obama established the first-ever White House Cancer Moonshot Task Force and committed $1 billion to accelerate research efforts against cancer and “bring about a decade’s worth of advances in five years.”
So, after combing through President Obama’s physical exam summaries from his time in the White House (from his first in 2010 to his most recent in 2016), why is there no mention of screening for lung cancer? Surely this decision can’t be due to rationing of healthcare resources or inadequate care by his personal physicians. How could everyone have ignored the specter of this lethal disease, given the unlimited resources for caring for the President and the potential to reduce his chances of dying from lung cancer by one-third? If you are puzzled, or would have cared for him differently, then perhaps the decision warrants a closer look.
Excerpts from President Obama’s most recent physical examination. Despite references to his smoking history and tobacco cessation efforts, no mention is made of possible lung cancer risk.
Early trials using chest x-rays to screen for lung cancer did not reduce lung cancer deaths, and actually seem to have increased the chance of death – likely due to overdiagnosis of innocuous nodules and the harms associated with more lung cancer biopsies and surgeries. The National Lung Screening Trial (NLST) in the early 2000s then randomly assigned over 50,000 smokers ages 55-74 to either annual low-dose computed tomography (CT), which combines x-rays from many different angles to produce detailed images of the body, or single chest x-ray to screen for lung cancer. All participants had accumulated at least 30 “pack-years,” a metric multiplying “years smoked” and “average packs-per-day smoked’ (such as someone who had smoked one pack-per-day for 30 years or two packs-per-day for 15 years). In the end, after a 6.5 year follow-up, the trial found 247 deaths from lung cancer per 100,000 person-years in the low-dose CT group and 309 deaths per 100,000 person-years in the radiography (x-ray) group. This represents a relative reduction in mortality from lung cancer with low-dose CT screening of 20.0%. (There were 62 fewer deaths per 100,000 person-years, and 62/309 is 20.0% of the deaths in the radiology group.) The trial has led the United States Preventive Services Task Force (USPSTF) to recommend annual low-dose CT for adults 55-80 years old who have a 30 pack-year smoking history and have smoked within the past 15 years.
President Obama probably doesn’t fall within these parameters but he certainly has some risk for lung cancer. By his own admission, he started smoking in his late teens but leading up to his presidential campaign in 2007 he had cut down to a few cigarettes a day (paling in comparison to Dwight Eisenhower’s reported four packs per day). His health report in 2008 noted that he had smoked intermittently, quitting on several occasions and at the time was using Nicorette gum with success. All together, he smoked on-and-off for roughly 30 years, likely averaging less than a pack per day. So what might his risk of lung cancer be?
Using data from the National Lung Screening Trial, a group at University of Michigan has developed and evaluated an online tool, shouldiscreen.com, to educate patients and predict personalized lung cancer risk. President Obama was 47 years old at the time of his inauguration in 2009, with possibly a 15 pack-year smoking history (30 years of 1/2 pack per day, admittedly a rough estimate). The rest of the risk calculator can be completed with information from his physicals exams – a height of 6’ 1”, weight of 175 lbs, a law degree, African-American race, no personal history of cancer or chronic pulmonary obstructive disease (COPD), and no family history of lung cancer.
The calculator reminds us that President Obama does not fit the criteria advocated by USPSTF for screening (ages 55-80 with a 30 pack-year smoking history) but extrapolates from NLST outcomes to estimate that, given his age and smoking history and other factors, his risk of developing lung cancer over a 6 year period is low at 0.3%. Though not perfectly generalizable, the tool does posit that out of 1,000 similar patients who were not screened, three would die from lung cancer. With complete low-dose CT screening there would still be two deaths from lung cancer, but compared to the original group, “there will be 1 fewer deaths out of 1000 in the next 6 years if you get screened.” So, reducing the chance of dying from lung cancer from 3/1000 to 2/1000 is a one-third reduction, or 33% decrease in risk. Why wouldn’t that be desirable?
It is helpful here to make a distinction between an “absolute risk” difference and a “relative risk” difference. Relative risk is obtained by division – since the new risk is 2/1000 and the original was 3/1000, it is 2/3 of the original (or, in other words, risk has decreased by 1/3). This is a 33% relative risk reduction. Absolute risk difference is derived by subtracting to find, in absolute terms (without reference to any original figure) how much the risk has changed. The absolute risk difference is 3/1000 – 2/1000, or 1/1000. The absolute change in risk is 0.1%. Does that sound less significant than a 33% [relative] risk reduction?
A “doubling of risk” refers to relative risk, and encompasses scenarios with very different changes in absolute risk. 1% → 2% and 10% → 20% are two examples. When a study describes a “33% risk reduction,” one has to ask if this is a relative risk reduction or an absolute risk reduction.
An absolute change in risk of 1 death/1,000 people means that 1,000 people must be screened to avert 1 death (this inversion of the absolute change in risk is known as the number needed to treat, or number needed to harm, depending on the direction of the effect). When 1,000 people are screened in this case, 1 is “treated” (their life is saved). And yet all 1,000 people who are screened are potentially subject to the harms of the screening test and subsequent diagnostic and therapeutic steps. This fact is also highlighted by shouldiscreen.com, which displays all of the harms alongside the potential benefits. In President Obama’s cohort of 1,000 similar patients, not only would 365 have non-cancerous nodules detected, but 18 would undergo invasive procedures and 3 would have a major complication. These figures are derived from the NLST where, perhaps not surprisingly, 96.4% of the positive screening results in the low-dose CT group were false positive results.
An output from shouldiscreen.com, which estimates the benefits and harms should a population of 1,000 people, similar in risk factors to Barack Obama, all be screened by annual low-dose computed tomography for lung cancer.
It is clear then why screening President Obama with low-dose CT for lung cancer would have been a risky endeavor. The 33% relative risk reduction trumpeted initially is really a 0.1% absolute reduction in the risk of death, meaning that President Obama would have perhaps the same chance of dying from an invasive procedure (0.3% had major complications) as he would have of averting a death from lung cancer.
And, incidentally, even an uncomplicated procedure has potential consequences when you’re the President of the United States. Did you know that Section 3 of the 25th Amendment, relating to the President’s ability to carry out their duties, has been used to transfer executive authority to the Vice President in circumstances such as when a President undergoes anesthesia? It has been implemented only three times – when President Ronald Reagan underwent colorectal cancer surgery, and twice when President George W. Bush underwent routine colonoscopies. The medical press releases for President Obama even explicitly state that such transfers were not necessary as he was never sedated for testing or procedures.
So, not unlike my mother’s screening exams, this is an instance is which more care would likely not have been better care. This is a unique discussion in that in no way do cost, resources, or access to healthcare play a role. President Obama, rightly so, has not been screened for lung cancer simply because such screening could cause him more harm, from further testing and invasive procedures, than good. A statement of relative risk reduction through screening (33%!) can be misleading but by examining the absolute risk reduction, personalized risk calculators, and possible harms associated with invasive diagnostic tests, we can make an evidence-based decision. And we see that we can care deeply about cancer and its outcomes without advocating for indiscriminate testing.
Do you have questions or comments on this article, or feedback for future columns in this epidemiology series? I welcome your thoughts at firstname.lastname@example.org!